PHXH
/
unity-mcp
51
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Merge commit '3e83f993bfe632034bf7302d4319e3cd16353eb8' into feat/telemetry 

* commit '3e83f993bfe632034bf7302d4319e3cd16353eb8':
  Improved ci prompt testing suite (#270)
  chore: bump version to 3.3.2
  Fix: Unity Editor reload crash + debug-noise reduction (#266)
  Revise README for improved clarity and organization
  docs: install uv via official installer (curl/winget)
  Update README.md
  docs: fix Windows uv path to use WinGet shim, keep macOS AppSupport symlink path
  docs: update README.md with improved installation paths, documentation, and logo
  fix: Update README installation paths to match ServerInstaller.cs
main
David Sarno 2025-09-08 10:11:46 -07:00
parent 99beca7cfe 3e83f993bf
commit 81dcd69722
27 changed files with 2102 additions and 1105 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

45
.claude/prompts/nl-unity-claude-tests-mini.md
View File

@ -1,45 +0,0 @@
# Unity NL Editing Suite — Natural Mode
You are running inside CI for the **unity-mcp** repository. Your task is to demonstrate end‑to‑end **natural‑language code editing** on a representative Unity C# script using whatever capabilities and servers are already available in this session. Work autonomously. Do not ask the user for input. Do NOT spawn subagents, as they will not have access to the mcp server process on the top-level agent.
## Mission
1) **Discover capabilities.** Quietly inspect the tools and any connected servers that are available to you at session start. If the server offers a primer or capabilities resource, read it before acting.
2) **Choose a target file.** Prefer `TestProjects/UnityMCPTests/Assets/Scripts/LongUnityScriptClaudeTest.cs` if it exists; otherwise choose a simple, safe C# script under `TestProjects/UnityMCPTests/Assets/`.
3) **Perform a small set of realistic edits** using minimal, precise changes (not full-file rewrites). Examples of small edits you may choose from (pick 3–6 total):
- Insert a new, small helper method (e.g., a logger or counter) in a sensible location.
- Add a short anchor comment near a key method (e.g., above `Update()`), then add or modify a few lines nearby.
- Append an end‑of‑class utility method (e.g., formatting or clamping helper).
- Make a safe, localized tweak to an existing method body (e.g., add a guard or a simple accumulator).
- Optionally include one idempotency/no‑op check (re‑apply an edit and confirm nothing breaks).
4) **Validate your edits.** Re‑read the modified regions and verify the changes exist, compile‑risk is low, and surrounding structure remains intact.
5) **Report results.** Produce both:
- A JUnit XML at `reports/junit-nl-suite.xml` containing a single suite named `UnityMCP.NL` with one test case per sub‑test you executed (mark pass/fail and include helpful failure text).
- A summary markdown at `reports/junit-nl-suite.md` that explains what you attempted, what succeeded/failed, and any follow‑ups you would try.
6) **Be gentle and reversible.** Prefer targeted, minimal edits; avoid wide refactors or non‑deterministic changes.
## Assumptions & Hints (non‑prescriptive)
- A Unity‑oriented MCP server is expected to be connected. If a server‑provided **primer/capabilities** resource exists, read it first. If no primer is available, infer capabilities from your visible tools in the session.
- In CI/headless mode, when calling `mcp__unity__list_resources` or `mcp__unity__read_resource`, include:
- `ctx: {}`
- `project_root: "TestProjects/UnityMCPTests"` (the server will also accept the absolute path passed via env)
Example: `{ "ctx": {}, "under": "Assets/Scripts", "pattern": "*.cs", "project_root": "TestProjects/UnityMCPTests" }`
- If the preferred file isn’t present, locate a fallback C# file with simple, local methods you can edit safely.
- If a compile command is available in this environment, you may optionally trigger it; if not, rely on structural checks and localized validation.
## Output Requirements (match NL suite conventions)
- JUnit XML at `$JUNIT_OUT` if set, otherwise `reports/junit-nl-suite.xml`.
- Single suite named `UnityMCP.NL`, one `<testcase>` per sub‑test; include `<failure>` on errors.
- Markdown at `$MD_OUT` if set, otherwise `reports/junit-nl-suite.md`.
Constraints (for fast publishing):
- Log allowed tools once as a single line: `AllowedTools: ...`.
- For every edit: Read → Write (with precondition hash) → Re‑read; on `{status:"stale_file"}` retry once after re‑read.
- Keep evidence to ±20–40 lines windows; cap unified diffs to 300 lines and note truncation.
- End `<system-out>` with `VERDICT: PASS` or `VERDICT: FAIL`.
## Guardrails
- No destructive operations. Keep changes minimal and well‑scoped.
- Don’t leak secrets or environment details beyond what’s needed in the reports.
- Work without user interaction; do not prompt for approval mid‑flow.
> If capabilities discovery fails, still produce the two reports that clearly explain why you could not proceed and what evidence you gathered.

234
.claude/prompts/nl-unity-suite-full.md
View File

@ -1,234 +0,0 @@
# Unity NL/T Editing Suite — CI Agent Contract
You are running inside CI for the `unity-mcp` repo. Use only the tools allowed by the workflow. Work autonomously; do not prompt the user. Do NOT spawn subagents.
**Print this once, verbatim, early in the run:**
AllowedTools: Write,Bash(printf:*),Bash(echo:*),Bash(scripts/nlt-revert.sh:*),mcp__unity__manage_editor,mcp__unity__list_resources,mcp__unity__read_resource,mcp__unity__apply_text_edits,mcp__unity__script_apply_edits,mcp__unity__validate_script,mcp__unity__find_in_file,mcp__unity__read_console,mcp__unity__get_sha
---
## Mission
1) Pick target file (prefer):
- `unity://path/Assets/Scripts/LongUnityScriptClaudeTest.cs`
2) Execute **all** NL/T tests in order using minimal, precise edits.
3) Validate each edit with `mcp__unity__validate_script(level:"standard")`.
4) **Report**: write one `<testcase>` XML fragment per test to `reports/<TESTID>_results.xml`. Do **not** read or edit `$JUNIT_OUT`.
5) **Restore** the file after each test using the OS‑level helper (fast), not a full‑file text write.
---
## Environment & Paths (CI)
- Always pass: `project_root: "TestProjects/UnityMCPTests"` and `ctx: {}` on list/read/edit/validate.
- **Canonical URIs only**:
- Primary: `unity://path/Assets/...` (never embed `project_root` in the URI)
- Relative (when supported): `Assets/...`
- File paths for the helper script are workspace‑relative:
- `TestProjects/UnityMCPTests/Assets/...`
CI provides:
- `$JUNIT_OUT=reports/junit-nl-suite.xml` (pre‑created; leave alone)
- `$MD_OUT=reports/junit-nl-suite.md` (synthesized from JUnit)
- Helper script: `scripts/nlt-revert.sh` (snapshot/restore)
---
## Tool Mapping
- **Anchors/regex/structured**: `mcp__unity__script_apply_edits`
- Allowed ops: `anchor_insert`, `replace_range`, `regex_replace` (no overlapping ranges within a single call)
- **Precise ranges / atomic batch**: `mcp__unity__apply_text_edits` (non‑overlapping ranges)
- Multi‑span batches are computed from the same fresh read and sent atomically by default.
- Prefer `options.applyMode:"atomic"` when passing options for multiple spans; for single‑span, sequential is fine.
- **Hash-only**: `mcp__unity__get_sha` — returns `{sha256,lengthBytes,lastModifiedUtc}` without file body
- **Validation**: `mcp__unity__validate_script(level:"standard")`
- For edits, you may pass `options.validate`:
- `standard` (default): full‑file delimiter balance checks.
- `relaxed`: scoped checks for interior, non‑structural text edits; do not use for header/signature/brace‑touching changes.
- **Reporting**: `Write` small XML fragments to `reports/*_results.xml`
- **Editor state/flush**: `mcp__unity__manage_editor` (use sparingly; no project mutations)
- **Console readback**: `mcp__unity__read_console` (INFO capture only; do not assert in place of `validate_script`)
- **Snapshot/Restore**: `Bash(scripts/nlt-revert.sh:*)`
- For `script_apply_edits`: use `name` + workspace‑relative `path` only (e.g., `name="LongUnityScriptClaudeTest"`, `path="Assets/Scripts"`). Do not pass `unity://...` URIs as `path`.
- For `apply_text_edits` / `read_resource`: use the URI form only (e.g., `uri="unity://path/Assets/Scripts/LongUnityScriptClaudeTest.cs"`). Do not concatenate `Assets/` with a `unity://...` URI.
- Never call generic Bash like `mkdir`; the revert helper creates needed directories. Use only `scripts/nlt-revert.sh` for snapshot/restore.
- If you believe a directory is missing, you are mistaken: the workflow pre-creates it and the snapshot helper creates it if needed. Do not attempt any Bash other than scripts/nlt-revert.sh:*.
### Structured edit ops (required usage)
# Insert a helper RIGHT BEFORE the final class brace (NL‑3, T‑D)
1) Prefer `script_apply_edits` with a regex capture on the final closing brace:
```json
{"op":"regex_replace",
"pattern":"(?s)(\\r?\\n\\s*\\})\\s*$",
"replacement":"\\n // Tail test A\\n // Tail test B\\n // Tail test C\\1"}
2) If the server returns `unsupported` (op not available) or `missing_field` (op‑specific), FALL BACK to
`apply_text_edits`:
- Find the last `}` in the file (class closing brace) by scanning from end.
- Insert the three comment lines immediately before that index with one non‑overlapping range.
# Insert after GetCurrentTarget (T‑A/T‑E)
- Use `script_apply_edits` with:
```json
{"op":"anchor_insert","afterMethodName":"GetCurrentTarget","text":"private int __TempHelper(int a,int b)=>a+b;\\n"}
```
# Delete the temporary helper (T‑A/T‑E)
- Prefer structured delete:
- Use `script_apply_edits` with `{ "op":"delete_method", "className":"LongUnityScriptClaudeTest", "methodName":"PrintSeries" }` (or `__TempHelper` for T‑A).
- If structured delete is unavailable, fall back to `apply_text_edits` with a single `replace_range` spanning the exact method block (bounds computed from a fresh read); avoid whole‑file regex deletes.
# T‑B (replace method body)
- Use `mcp__unity__apply_text_edits` with a single `replace_range` strictly inside the `HasTarget` braces.
- Compute start/end from a fresh `read_resource` at test start. Do not edit signature or header.
- On `{status:"stale_file"}` retry once with the server-provided hash; if absent, re-read once and retry.
- On `bad_request`: write the testcase with `<failure>…</failure>`, restore, and continue to next test.
- On `missing_field`: FALL BACK per above; if the fallback also returns `unsupported` or `bad_request`, then fail as above.
> Don’t use `mcp__unity__create_script`. Avoid the header/`using` region entirely.
Span formats for `apply_text_edits`:
- Prefer LSP ranges (0‑based): `{ "range": { "start": {"line": L, "character": C}, "end": {…} }, "newText": "…" }`
- Explicit fields are 1‑based: `{ "startLine": L1, "startCol": C1, "endLine": L2, "endCol": C2, "newText": "…" }`
- SDK preflights overlap after normalization; overlapping non‑zero spans → `{status:"overlap"}` with conflicts and no file mutation.
- Optional debug: pass `strict:true` to reject explicit 0‑based fields (else they are normalized and a warning is emitted).
- Apply mode guidance: router defaults to atomic for multi‑span; you can explicitly set `options.applyMode` if needed.
---
## Output Rules (JUnit fragments only)
- For each test, create **one** file: `reports/<TESTID>_results.xml` containing exactly a single `<testcase ...> ... </testcase>`.
Put human-readable lines (PLAN/PROGRESS/evidence) **inside** `<system-out><![CDATA[ ... ]]></system-out>`.
- If content contains `]]>`, split CDATA: replace `]]>` with `]]]]><![CDATA[>`.
- Evidence windows only (±20–40 lines). If showing a unified diff, cap at 100 lines and note truncation.
- **Never** open/patch `$JUNIT_OUT` or `$MD_OUT`; CI merges fragments and synthesizes Markdown.
- Write destinations must match: `^reports/[A-Za-z0-9._-]+_results\.xml$`
- Snapshot files must live under `reports/_snapshots/`
- Reject absolute paths and any path containing `..`
- Reject control characters and line breaks in filenames; enforce UTF‑8
- Cap basename length to ≤64 chars; cap any path segment to ≤100 and total path length to ≤255
- Bash(printf|echo) must write to stdout only. Do not use shell redirection, here‑docs, or `tee` to create/modify files. The only allowed FS mutation is via `scripts/nlt-revert.sh`.
**Example fragment**
```xml
<testcase classname="UnityMCP.NL-T" name="NL-1. Method replace/insert/delete">
<system-out><![CDATA[
PLAN: NL-0,NL-1,NL-2,NL-3,NL-4,T-A,T-B,T-C,T-D,T-E,T-F,T-G,T-H,T-I,T-J (len=15)
PROGRESS: 2/15 completed
pre_sha=<...>
... evidence windows ...
VERDICT: PASS
]]></system-out>
</testcase>
```
Note: Emit the PLAN line only in NL‑0 (do not repeat it for later tests).
### Fast Restore Strategy (OS‑level)
- Snapshot once at NL‑0, then restore after each test via the helper.
- Snapshot (once after confirming the target):
```bash
scripts/nlt-revert.sh snapshot "TestProjects/UnityMCPTests/Assets/Scripts/LongUnityScriptClaudeTest.cs" "reports/_snapshots/LongUnityScriptClaudeTest.cs.baseline"
```
- Log `snapshot_sha=...` printed by the script.
- Restore (after each mutating test):
```bash
scripts/nlt-revert.sh restore "TestProjects/UnityMCPTests/Assets/Scripts/LongUnityScriptClaudeTest.cs" "reports/_snapshots/LongUnityScriptClaudeTest.cs.baseline"
```
- Then `read_resource` to confirm and (optionally) `validate_script(level:"standard")`.
- If the helper fails: fall back once to a guarded full‑file restore using the baseline bytes; then continue.
### Guarded Write Pattern (for edits, not restores)
- Before any mutation: `res = mcp__unity__read_resource(uri)`; `pre_sha = sha256(res.bytes)`.
- Write with `precondition_sha256 = pre_sha` on `apply_text_edits`/`script_apply_edits`.
- To compute `pre_sha` without reading file contents, you may instead call `mcp__unity__get_sha(uri).sha256`.
- On `{status:"stale_file"}`:
- Retry once using the server-provided hash (e.g., `data.current_sha256` or `data.expected_sha256`, per API schema).
- If absent, one re-read then a final retry. No loops.
- After success: immediately re-read via `res2 = mcp__unity__read_resource(uri)` and set `pre_sha = sha256(res2.bytes)` before any further edits in the same test.
- Prefer anchors (`script_apply_edits`) for end-of-class / above-method insertions. Keep edits inside method bodies. Avoid header/using.
**On non‑JSON/transport errors (timeout, EOF, connection closed):**
- Write `reports/<TESTID>_results.xml` with a `<testcase>` that includes a `<failure>` or `<error>` node capturing the error text.
- Run the OS restore via `scripts/nlt-revert.sh restore …`.
- Continue to the next test (do not abort).
**If any write returns `bad_request`, or `unsupported` after a fallback attempt:**
- Write `reports/<TESTID>_results.xml` with a `<testcase>` that includes a `<failure>` node capturing the server error, include evidence, and end with `VERDICT: FAIL`.
- Run `scripts/nlt-revert.sh restore ...` and continue to the next test.
### Execution Order (fixed)
- Run exactly: NL-0, NL-1, NL-2, NL-3, NL-4, T-A, T-B, T-C, T-D, T-E, T-F, T-G, T-H, T-I, T-J (15 total).
- Before NL-1..T-J: Bash(scripts/nlt-revert.sh:restore "<target>" "reports/_snapshots/LongUnityScriptClaudeTest.cs.baseline") IF the baseline exists; skip for NL-0.
- NL-0 must include the PLAN line (len=15).
- After each testcase, include `PROGRESS: <k>/15 completed`.
### Test Specs (concise)
- NL‑0. Sanity reads — Tail ~120; ±40 around `Update()`. Then snapshot via helper.
- NL‑1. Replace/insert/delete — `HasTarget → return currentTarget != null;`; insert `PrintSeries()` after `GetCurrentTarget` logging "1,2,3"; verify; delete `PrintSeries()`; restore.
- NL‑2. Anchor comment — Insert `// Build marker OK` above `public void Update(...)`; restore.
- NL‑3. End‑of‑class — Insert `// Tail test A/B/C` (3 lines) before final brace; restore.
- NL‑4. Compile trigger — Record INFO only.
### T‑A. Anchor insert (text path) — Insert helper after `GetCurrentTarget`; verify; delete via `regex_replace`; restore.
### T‑B. Replace body — Single `replace_range` inside `HasTarget`; restore.
- Options: pass {"validate":"relaxed"} for interior one-line edits.
### T‑C. Header/region preservation — Edit interior of `ApplyBlend`; preserve signature/docs/regions; restore.
- Options: pass {"validate":"relaxed"} for interior one-line edits.
### T‑D. End‑of‑class (anchor) — Insert helper before final brace; remove; restore.
### T‑E. Lifecycle — Insert → update → delete via regex; restore.
### T‑F. Atomic batch — One `mcp__unity__apply_text_edits` call (text ranges only)
- Compute all three edits from the **same fresh read**:
1) Two small interior `replace_range` tweaks.
2) One **end‑of‑class insertion**: find the **index of the final `}`** for the class; create a zero‑width range `[idx, idx)` and set `replacement` to the 3‑line comment block.
- Send all three ranges in **one call**, sorted **descending by start index** to avoid offset drift.
- Expect all‑or‑nothing semantics; on `{status:"overlap"}` or `{status:"bad_request"}`, write the testcase fragment with `<failure>…</failure>`, **restore**, and continue.
- Options: pass {"applyMode":"atomic"} to enforce all‑or‑nothing.
- T‑G. Path normalization — Make the same edit with `unity://path/Assets/...` then `Assets/...`. Without refreshing `precondition_sha256`, the second attempt returns `{stale_file}`; retry with the server-provided hash to confirm both forms resolve to the same file.
### T-H. Validation (standard)
- Restore baseline (helper call above).
- Perform a harmless interior tweak (or none), then MUST call:
mcp__unity__validate_script(level:"standard")
- Write the validator output to system-out; VERDICT: PASS if standard is clean, else include <failure> with the validator message and continue.
### T-I. Failure surfaces (expected)
- Restore baseline.
- (1) OVERLAP:
* Fresh read of file; compute two interior ranges that overlap inside HasTarget.
* Prefer LSP ranges (0‑based) or explicit 1‑based fields; ensure both spans come from the same snapshot.
* Single mcp__unity__apply_text_edits call with both ranges.
* Expect `{status:"overlap"}` (SDK preflight) → record as PASS; else FAIL. Restore.
- (2) STALE_FILE:
* Fresh read → pre_sha.
* Make a tiny legit edit with pre_sha; success.
* Attempt another edit reusing the OLD pre_sha.
* Expect {status:"stale_file"} → record as PASS; else FAIL. Re-read to refresh, restore.
### Per‑test error handling and recovery
- For each test (NL‑0..T‑J), use a try/finally pattern:
- Always write a testcase fragment and perform restore in finally, even when tools return error payloads.
- try: run the test steps; always write `reports/<ID>_results.xml` with PASS/FAIL/ERROR
- finally: run Bash(scripts/nlt-revert.sh:restore …baseline) to restore the target file
- On any transport/JSON/tool exception:
- catch and write a `<testcase>` fragment with an `<error>` node (include the message), then proceed to the next test.
- After NL‑4 completes, proceed directly to T‑A regardless of any earlier validator warnings (do not abort the run).
- (3) USING_GUARD (optional):
* Attempt a 1-line insert above the first 'using'.
* Expect {status:"using_guard"} → record as PASS; else note 'not emitted'. Restore.
### T-J. Idempotency
- Restore baseline.
- Repeat a replace_range twice (second call may be noop). Validate standard after each.
- Insert or ensure a tiny comment, then delete it twice (second delete may be noop).
- Restore and PASS unless an error/structural break occurred.
### Status & Reporting
- Safeguard statuses are non‑fatal; record and continue.
- End each testcase `<system-out>` with `VERDICT: PASS` or `VERDICT: FAIL`.

144
.claude/prompts/nl-unity-suite-full-additive.md → .claude/prompts/nl-unity-suite-nl.md
View File

@ -1,4 +1,4 @@
# Unity NL/T Editing Suite — Additive Test Design # Unity NL Editing Suite — Additive Test Design
You are running inside CI for the `unity-mcp` repo. Use only the tools allowed by the workflow. Work autonomously; do not prompt the user. Do NOT spawn subagents. You are running inside CI for the `unity-mcp` repo. Use only the tools allowed by the workflow. Work autonomously; do not prompt the user. Do NOT spawn subagents.
@ -10,10 +10,28 @@ AllowedTools: Write,mcp__unity__manage_editor,mcp__unity__list_resources,mcp__un
## Mission ## Mission
1) Pick target file (prefer): 1) Pick target file (prefer):
- `unity://path/Assets/Scripts/LongUnityScriptClaudeTest.cs` - `unity://path/Assets/Scripts/LongUnityScriptClaudeTest.cs`
2) Execute **all** NL/T tests in order using minimal, precise edits that **build on each other**. 2) Execute NL tests NL-0..NL-4 in order using minimal, precise edits that build on each other.
3) Validate each edit with `mcp__unity__validate_script(level:"standard")`. 3) Validate each edit with `mcp__unity__validate_script(level:"standard")`.
4) **Report**: write one `<testcase>` XML fragment per test to `reports/<TESTID>_results.xml`. Do **not** read or edit `$JUNIT_OUT`. 4) **Report**: write one `<testcase>` XML fragment per test to `reports/<TESTID>_results.xml`. Do **not** read or edit `$JUNIT_OUT`.
**CRITICAL XML FORMAT REQUIREMENTS:**
- Each file must contain EXACTLY one `<testcase>` root element
- NO prologue, epilogue, code fences, or extra characters
- NO markdown formatting or explanations outside the XML
- Use this exact format:
```xml
<testcase name="NL-0 — Baseline State Capture" classname="UnityMCP.NL-T">
<system-out><![CDATA[
(evidence of what was accomplished)
]]></system-out>
</testcase>
```
- If test fails, include: `<failure message="reason"/>`
- TESTID must be one of: NL-0, NL-1, NL-2, NL-3, NL-4
5) **NO RESTORATION** - tests build additively on previous state. 5) **NO RESTORATION** - tests build additively on previous state.
6) **STRICT FRAGMENT EMISSION** - After each test, immediately emit a clean XML file under `reports/<TESTID>_results.xml` with exactly one `<testcase>` whose `name` begins with the exact test id. No prologue/epilogue or fences. If the test fails, include a `<failure message="..."/>` and still emit.
--- ---
@ -29,10 +47,26 @@ CI provides:
--- ---
## Transcript Minimization Rules
- Do not restate tool JSON; summarize in ≤ 2 short lines.
- Never paste full file contents. For matches, include only the matched line and ±1 line.
- Prefer `mcp__unity__find_in_file` for targeting; avoid `mcp__unity__read_resource` unless strictly necessary. If needed, limit to `head_bytes ≤ 256` or `tail_lines ≤ 10`.
- Per‑test `system-out` ≤ 400 chars: brief status only (no SHA).
- Console evidence: fetch the last 10 lines with `include_stacktrace:false` and include ≤ 3 lines in the fragment.
- Avoid quoting multi‑line diffs; reference markers instead.
— Console scans: perform two reads — last 10 `log/info` lines and up to 3 `error` entries (use `include_stacktrace:false`); include ≤ 3 lines total in the fragment; if no errors, state "no errors".
---
## Tool Mapping ## Tool Mapping
- **Anchors/regex/structured**: `mcp__unity__script_apply_edits` - **Anchors/regex/structured**: `mcp__unity__script_apply_edits`
- Allowed ops: `anchor_insert`, `replace_method`, `insert_method`, `delete_method`, `regex_replace` - Allowed ops: `anchor_insert`, `replace_method`, `insert_method`, `delete_method`, `regex_replace`
- For `anchor_insert`, always set `"position": "before"` or `"after"`.
- **Precise ranges / atomic batch**: `mcp__unity__apply_text_edits` (non‑overlapping ranges) - **Precise ranges / atomic batch**: `mcp__unity__apply_text_edits` (non‑overlapping ranges)
STRICT OP GUARDRAILS
- Do not use `anchor_replace`. Structured edits must be one of: `anchor_insert`, `replace_method`, `insert_method`, `delete_method`, `regex_replace`.
- For multi‑spot textual tweaks in one operation, compute non‑overlapping ranges with `mcp__unity__find_in_file` and use `mcp__unity__apply_text_edits`.
- **Hash-only**: `mcp__unity__get_sha` — returns `{sha256,lengthBytes,lastModifiedUtc}` without file body - **Hash-only**: `mcp__unity__get_sha` — returns `{sha256,lengthBytes,lastModifiedUtc}` without file body
- **Validation**: `mcp__unity__validate_script(level:"standard")` - **Validation**: `mcp__unity__validate_script(level:"standard")`
- **Dynamic targeting**: Use `mcp__unity__find_in_file` to locate current positions of methods/markers - **Dynamic targeting**: Use `mcp__unity__find_in_file` to locate current positions of methods/markers
@ -49,7 +83,7 @@ CI provides:
5. **Composability**: Tests demonstrate how operations work together in real workflows 5. **Composability**: Tests demonstrate how operations work together in real workflows
**State Tracking:** **State Tracking:**
- Track file SHA after each test to ensure operations succeeded - Track file SHA after each test (`mcp__unity__get_sha`) for potential preconditions in later passes. Do not include SHA values in report fragments.
- Use content signatures (method names, comment markers) to verify expected state - Use content signatures (method names, comment markers) to verify expected state
- Validate structural integrity after each major change - Validate structural integrity after each major change
@ -85,7 +119,8 @@ CI provides:
### NL-3. End-of-Class Content (Additive State C) ### NL-3. End-of-Class Content (Additive State C)
**Goal**: Demonstrate end-of-class insertions with smart brace matching **Goal**: Demonstrate end-of-class insertions with smart brace matching
**Actions**: **Actions**:
- Use anchor pattern to find the class-ending brace (accounts for previous additions) - Match the final class-closing brace by scanning from EOF (e.g., last `^\s*}\s*$`)
or compute via `find_in_file` + ranges; insert immediately before it.
- Insert three comment lines before final class brace: - Insert three comment lines before final class brace:
``` ```
// Tail test A // Tail test A
@ -97,95 +132,11 @@ CI provides:
### NL-4. Console State Verification (No State Change) ### NL-4. Console State Verification (No State Change)
**Goal**: Verify Unity console integration without file modification **Goal**: Verify Unity console integration without file modification
**Actions**: **Actions**:
- Read Unity console messages (INFO level) - Read last 10 Unity console lines (log/info)
- Perform a targeted scan for errors/exceptions (type: errors), up to 3 entries
- Validate no compilation errors from previous operations - Validate no compilation errors from previous operations
- **Expected final state**: State C (unchanged) - **Expected final state**: State C (unchanged)
- **IMMEDIATELY** write clean XML fragment to `reports/NL-4_results.xml` (no extra text). The `<testcase name>` must start with `NL-4`. Include at most 3 lines total across both reads, or simply state "no errors; console OK" (≤ 400 chars).
### T-A. Temporary Helper Lifecycle (Returns to State C)
**Goal**: Test insert → verify → delete cycle for temporary code
**Actions**:
- Find current position of `GetCurrentTarget()` method (may have shifted from NL-2 comment)
- Insert temporary helper: `private int __TempHelper(int a, int b) => a + b;`
- Verify helper method exists and compiles
- Delete helper method via structured delete operation
- **Expected final state**: Return to State C (helper removed, other changes intact)
### T-B. Method Body Interior Edit (Additive State D)
**Goal**: Edit method interior without affecting structure, on modified file
**Actions**:
- Use `find_in_file` to locate current `HasTarget()` method (modified in NL-1)
- Edit method body interior: change return statement to `return true; /* test modification */`
- Use `validate: "relaxed"` for interior-only edit
- Verify edit succeeded and file remains balanced
- **Expected final state**: State C + modified HasTarget() body
### T-C. Different Method Interior Edit (Additive State E)
**Goal**: Edit a different method to show operations don't interfere
**Actions**:
- Locate `ApplyBlend()` method using content search
- Edit interior line to add null check: `if (animator == null) return; // safety check`
- Preserve method signature and structure
- **Expected final state**: State D + modified ApplyBlend() method
### T-D. End-of-Class Helper (Additive State F)
**Goal**: Add permanent helper method at class end
**Actions**:
- Use smart anchor matching to find current class-ending brace (after NL-3 tail comments)
- Insert permanent helper before class brace: `private void TestHelper() { /* placeholder */ }`
- **Expected final state**: State E + TestHelper() method before class end
### T-E. Method Evolution Lifecycle (Additive State G)
**Goal**: Insert → modify → finalize a method through multiple operations
**Actions**:
- Insert basic method: `private int Counter = 0;`
- Update it: find and replace with `private int Counter = 42; // initialized`
- Add companion method: `private void IncrementCounter() { Counter++; }`
- **Expected final state**: State F + Counter field + IncrementCounter() method
### T-F. Atomic Multi-Edit (Additive State H)
**Goal**: Multiple coordinated edits in single atomic operation
**Actions**:
- Read current file state to compute precise ranges
- Atomic edit combining:
1. Add comment in `HasTarget()`: `// validated access`
2. Add comment in `ApplyBlend()`: `// safe animation`
3. Add final class comment: `// end of test modifications`
- All edits computed from same file snapshot, applied atomically
- **Expected final state**: State G + three coordinated comments
### T-G. Path Normalization Test (No State Change)
**Goal**: Verify URI forms work equivalently on modified file
**Actions**:
- Make identical edit using `unity://path/Assets/Scripts/LongUnityScriptClaudeTest.cs`
- Then using `Assets/Scripts/LongUnityScriptClaudeTest.cs`
- Second should return `stale_file`, retry with updated SHA
- Verify both URI forms target same file
- **Expected final state**: State H (no content change, just path testing)
### T-H. Validation on Modified File (No State Change)
**Goal**: Ensure validation works correctly on heavily modified file
**Actions**:
- Run `validate_script(level:"standard")` on current state
- Verify no structural errors despite extensive modifications
- **Expected final state**: State H (validation only, no edits)
### T-I. Failure Surface Testing (No State Change)
**Goal**: Test error handling on real modified file
**Actions**:
- Attempt overlapping edits (should fail cleanly)
- Attempt edit with stale SHA (should fail cleanly)
- Verify error responses are informative
- **Expected final state**: State H (failed operations don't modify file)
### T-J. Idempotency on Modified File (Additive State I)
**Goal**: Verify operations behave predictably when repeated
**Actions**:
- Add unique marker comment: `// idempotency test marker`
- Attempt to add same comment again (should detect no-op)
- Remove marker, attempt removal again (should handle gracefully)
- **Expected final state**: State H + verified idempotent behavior
---
## Dynamic Targeting Examples ## Dynamic Targeting Examples
@ -219,7 +170,8 @@ find_in_file(pattern: "public bool HasTarget\\(\\)")
1. Verify expected content exists: `find_in_file` for key markers 1. Verify expected content exists: `find_in_file` for key markers
2. Check structural integrity: `validate_script(level:"standard")` 2. Check structural integrity: `validate_script(level:"standard")`
3. Update SHA tracking for next test's preconditions 3. Update SHA tracking for next test's preconditions
4. Log cumulative changes in test evidence 4. Emit a per‑test fragment to `reports/<TESTID>_results.xml` immediately. If the test failed, still write a single `<testcase>` with a `<failure message="..."/>` and evidence in `system-out`.
5. Log cumulative changes in test evidence (keep concise per Transcript Minimization Rules; never paste raw tool JSON)
**Error Recovery:** **Error Recovery:**
- If test fails, log current state but continue (don't restore) - If test fails, log current state but continue (don't restore)
@ -238,3 +190,11 @@ find_in_file(pattern: "public bool HasTarget\\(\\)")
6. **State Evolution Testing**: Validates SDK handles cumulative file modifications correctly 6. **State Evolution Testing**: Validates SDK handles cumulative file modifications correctly
This additive approach produces a more realistic and maintainable test suite that better represents actual SDK usage patterns. This additive approach produces a more realistic and maintainable test suite that better represents actual SDK usage patterns.
---
BAN ON EXTRA TOOLS AND DIRS
- Do not use any tools outside `AllowedTools`. Do not create directories; assume `reports/` exists.
---

305
.claude/prompts/nl-unity-suite-t.md Normal file
View File

@ -0,0 +1,305 @@
# Unity T Editing Suite — Additive Test Design
You are running inside CI for the `unity-mcp` repo. Use only the tools allowed by the workflow. Work autonomously; do not prompt the user. Do NOT spawn subagents.
**Print this once, verbatim, early in the run:**
AllowedTools: Write,mcp__unity__manage_editor,mcp__unity__list_resources,mcp__unity__read_resource,mcp__unity__apply_text_edits,mcp__unity__script_apply_edits,mcp__unity__validate_script,mcp__unity__find_in_file,mcp__unity__read_console,mcp__unity__get_sha
---
## Mission
1) Pick target file (prefer):
- `unity://path/Assets/Scripts/LongUnityScriptClaudeTest.cs`
2) Execute T tests T-A..T-J in order using minimal, precise edits that build on the NL pass state.
3) Validate each edit with `mcp__unity__validate_script(level:"standard")`.
4) **Report**: write one `<testcase>` XML fragment per test to `reports/<TESTID>_results.xml`. Do **not** read or edit `$JUNIT_OUT`.
**CRITICAL XML FORMAT REQUIREMENTS:**
- Each file must contain EXACTLY one `<testcase>` root element
- NO prologue, epilogue, code fences, or extra characters
- NO markdown formatting or explanations outside the XML
- Use this exact format:
```xml
<testcase name="T-D — End-of-Class Helper" classname="UnityMCP.NL-T">
<system-out><![CDATA[
(evidence of what was accomplished)
]]></system-out>
</testcase>
```
- If test fails, include: `<failure message="reason"/>`
- TESTID must be one of: T-A, T-B, T-C, T-D, T-E, T-F, T-G, T-H, T-I, T-J
5) **NO RESTORATION** - tests build additively on previous state.
6) **STRICT FRAGMENT EMISSION** - After each test, immediately emit a clean XML file under `reports/<TESTID>_results.xml` with exactly one `<testcase>` whose `name` begins with the exact test id. No prologue/epilogue or fences. If the test fails, include a `<failure message="..."/>` and still emit.
---
## Environment & Paths (CI)
- Always pass: `project_root: "TestProjects/UnityMCPTests"` and `ctx: {}` on list/read/edit/validate.
- **Canonical URIs only**:
- Primary: `unity://path/Assets/...` (never embed `project_root` in the URI)
- Relative (when supported): `Assets/...`
CI provides:
- `$JUNIT_OUT=reports/junit-nl-suite.xml` (pre‑created; leave alone)
- `$MD_OUT=reports/junit-nl-suite.md` (synthesized from JUnit)
---
## Transcript Minimization Rules
- Do not restate tool JSON; summarize in ≤ 2 short lines.
- Never paste full file contents. For matches, include only the matched line and ±1 line.
- Prefer `mcp__unity__find_in_file` for targeting; avoid `mcp__unity__read_resource` unless strictly necessary. If needed, limit to `head_bytes ≤ 256` or `tail_lines ≤ 10`.
- Per‑test `system-out` ≤ 400 chars: brief status only (no SHA).
- Console evidence: fetch the last 10 lines with `include_stacktrace:false` and include ≤ 3 lines in the fragment.
- Avoid quoting multi‑line diffs; reference markers instead.
— Console scans: perform two reads — last 10 `log/info` lines and up to 3 `error` entries (use `include_stacktrace:false`); include ≤ 3 lines total in the fragment; if no errors, state "no errors".
— Final check is folded into T‑J: perform an errors‑only scan (with `include_stacktrace:false`) and include a single "no errors" line or up to 3 error lines within the T‑J fragment.
---
## Tool Mapping
- **Anchors/regex/structured**: `mcp__unity__script_apply_edits`
- Allowed ops: `anchor_insert`, `replace_method`, `insert_method`, `delete_method`, `regex_replace`
- For `anchor_insert`, always set `"position": "before"` or `"after"`.
- **Precise ranges / atomic batch**: `mcp__unity__apply_text_edits` (non‑overlapping ranges)
STRICT OP GUARDRAILS
- Do not use `anchor_replace`. Structured edits must be one of: `anchor_insert`, `replace_method`, `insert_method`, `delete_method`, `regex_replace`.
- For multi‑spot textual tweaks in one operation, compute non‑overlapping ranges with `mcp__unity__find_in_file` and use `mcp__unity__apply_text_edits`.
- **Hash-only**: `mcp__unity__get_sha` — returns `{sha256,lengthBytes,lastModifiedUtc}` without file body
- **Validation**: `mcp__unity__validate_script(level:"standard")`
- **Dynamic targeting**: Use `mcp__unity__find_in_file` to locate current positions of methods/markers
---
## Additive Test Design Principles
**Key Changes from Reset-Based:**
1. **Dynamic Targeting**: Use `find_in_file` to locate methods/content, never hardcode line numbers
2. **State Awareness**: Each test expects the file state left by the previous test
3. **Content-Based Operations**: Target methods by signature, classes by name, not coordinates
4. **Cumulative Validation**: Ensure the file remains structurally sound throughout the sequence
5. **Composability**: Tests demonstrate how operations work together in real workflows
**State Tracking:**
- Track file SHA after each test (`mcp__unity__get_sha`) and use it as a precondition
for `apply_text_edits` in T‑F/T‑G/T‑I to exercise `stale_file` semantics. Do not include SHA values in report fragments.
- Use content signatures (method names, comment markers) to verify expected state
- Validate structural integrity after each major change
---
### T-A. Temporary Helper Lifecycle (Returns to State C)
**Goal**: Test insert → verify → delete cycle for temporary code
**Actions**:
- Find current position of `GetCurrentTarget()` method (may have shifted from NL-2 comment)
- Insert temporary helper: `private int __TempHelper(int a, int b) => a + b;`
- Verify helper method exists and compiles
- Delete helper method via structured delete operation
- **Expected final state**: Return to State C (helper removed, other changes intact)
### Late-Test Editing Rule
- When modifying a method body, use `mcp__unity__script_apply_edits`. If the method is expression-bodied (`=>`), convert it to a block or replace the whole method definition. After the edit, run `mcp__unity__validate_script` and rollback on error. Use `//` comments in inserted code.
### T-B. Method Body Interior Edit (Additive State D)
**Goal**: Edit method interior without affecting structure, on modified file
**Actions**:
- Use `find_in_file` to locate current `HasTarget()` method (modified in NL-1)
- Edit method body interior: change return statement to `return true; /* test modification */`
- Validate with `mcp__unity__validate_script(level:"standard")` for consistency
- Verify edit succeeded and file remains balanced
- **Expected final state**: State C + modified HasTarget() body
### T-C. Different Method Interior Edit (Additive State E)
**Goal**: Edit a different method to show operations don't interfere
**Actions**:
- Locate `ApplyBlend()` method using content search
- Edit interior line to add null check: `if (animator == null) return; // safety check`
- Preserve method signature and structure
- **Expected final state**: State D + modified ApplyBlend() method
### T-D. End-of-Class Helper (Additive State F)
**Goal**: Add permanent helper method at class end
**Actions**:
- Use smart anchor matching to find current class-ending brace (after NL-3 tail comments)
- Insert permanent helper before class brace: `private void TestHelper() { /* placeholder */ }`
- Validate with `mcp__unity__validate_script(level:"standard")`
- **IMMEDIATELY** write clean XML fragment to `reports/T-D_results.xml` (no extra text). The `<testcase name>` must start with `T-D`. Include brief evidence in `system-out`.
- **Expected final state**: State E + TestHelper() method before class end
### T-E. Method Evolution Lifecycle (Additive State G)
**Goal**: Insert → modify → finalize a field + companion method
**Actions**:
- Insert field: `private int Counter = 0;`
- Update it: find and replace with `private int Counter = 42; // initialized`
- Add companion method: `private void IncrementCounter() { Counter++; }`
- **Expected final state**: State F + Counter field + IncrementCounter() method
### T-F. Atomic Multi-Edit (Additive State H)
**Goal**: Multiple coordinated edits in single atomic operation
**Actions**:
- Read current file state to compute precise ranges
- Atomic edit combining:
1. Add comment in `HasTarget()`: `// validated access`
2. Add comment in `ApplyBlend()`: `// safe animation`
3. Add final class comment: `// end of test modifications`
- All edits computed from same file snapshot, applied atomically
- **Expected final state**: State G + three coordinated comments
- After applying the atomic edits, run `validate_script(level:"standard")` and emit a clean fragment to `reports/T-F_results.xml` with a short summary.
### T-G. Path Normalization Test (No State Change)
**Goal**: Verify URI forms work equivalently on modified file
**Actions**:
- Make identical edit using `unity://path/Assets/Scripts/LongUnityScriptClaudeTest.cs`
- Then using `Assets/Scripts/LongUnityScriptClaudeTest.cs`
- Second should return `stale_file`, retry with updated SHA
- Verify both URI forms target same file
- **Expected final state**: State H (no content change, just path testing)
- Emit `reports/T-G_results.xml` showing evidence of stale SHA handling.
### T-H. Validation on Modified File (No State Change)
**Goal**: Ensure validation works correctly on heavily modified file
**Actions**:
- Run `validate_script(level:"standard")` on current state
- Verify no structural errors despite extensive modifications
- **Expected final state**: State H (validation only, no edits)
- Emit `reports/T-H_results.xml` confirming validation OK.
### T-I. Failure Surface Testing (No State Change)
**Goal**: Test error handling on real modified file
**Actions**:
- Attempt overlapping edits (should fail cleanly)
- Attempt edit with stale SHA (should fail cleanly)
- Verify error responses are informative
- **Expected final state**: State H (failed operations don't modify file)
- Emit `reports/T-I_results.xml` capturing error evidence; file must contain one `<testcase>`.
### T-J. Idempotency on Modified File (Additive State I)
**Goal**: Verify operations behave predictably when repeated
**Actions**:
- **Insert (structured)**: `mcp__unity__script_apply_edits` with:
`{"op":"anchor_insert","anchor":"// Tail test C","position":"after","text":"\n // idempotency test marker"}`
- **Insert again** (same op) → expect `no_op: true`.
- **Remove (structured)**: `{"op":"regex_replace","pattern":"(?m)^\\s*// idempotency test marker\\r?\\n?","text":""}`
- **Remove again** (same `regex_replace`) → expect `no_op: true`.
- `mcp__unity__validate_script(level:"standard")`
- Perform a final console scan for errors/exceptions (errors only, up to 3); include "no errors" if none
- **IMMEDIATELY** write clean XML fragment to `reports/T-J_results.xml` with evidence of both `no_op: true` outcomes and the console result. The `<testcase name>` must start with `T-J`.
- **Expected final state**: State H + verified idempotent behavior
---
## Dynamic Targeting Examples
**Instead of hardcoded coordinates:**
```json
{"startLine": 31, "startCol": 26, "endLine": 31, "endCol": 58}
```
**Use content-aware targeting:**
```json
# Find current method location
find_in_file(pattern: "public bool HasTarget\\(\\)")
# Then compute edit ranges from found position
```
**Method targeting by signature:**
```json
{"op": "replace_method", "className": "LongUnityScriptClaudeTest", "methodName": "HasTarget"}
```
**Anchor-based insertions:**
```json
{"op": "anchor_insert", "anchor": "private void Update\\(\\)", "position": "before", "text": "// comment"}
```
---
## State Verification Patterns
**After each test:**
1. Verify expected content exists: `find_in_file` for key markers
2. Check structural integrity: `validate_script(level:"standard")`
3. Update SHA tracking for next test's preconditions
4. Emit a per‑test fragment to `reports/<TESTID>_results.xml` immediately. If the test failed, still write a single `<testcase>` with a `<failure message="..."/>` and evidence in `system-out`.
5. Log cumulative changes in test evidence (keep concise per Transcript Minimization Rules; never paste raw tool JSON)
**Error Recovery:**
- If test fails, log current state but continue (don't restore)
- Next test adapts to actual current state, not expected state
- Demonstrates resilience of operations on varied file conditions
---
## Benefits of Additive Design
1. **Realistic Workflows**: Tests mirror actual development patterns
2. **Robust Operations**: Proves edits work on evolving files, not just pristine baselines
3. **Composability Validation**: Shows operations coordinate well together
4. **Simplified Infrastructure**: No restore scripts or snapshots needed
5. **Better Failure Analysis**: Failures don't cascade - each test adapts to current reality
6. **State Evolution Testing**: Validates SDK handles cumulative file modifications correctly
This additive approach produces a more realistic and maintainable test suite that better represents actual SDK usage patterns.
---
BAN ON EXTRA TOOLS AND DIRS
- Do not use any tools outside `AllowedTools`. Do not create directories; assume `reports/` exists.
---
## XML Fragment Templates (T-F .. T-J)
Use these skeletons verbatim as a starting point. Replace the bracketed placeholders with your evidence. Ensure each file contains exactly one `<testcase>` element and that the `name` begins with the exact test id.
```xml
<testcase name="T-F — Atomic Multi-Edit" classname="UnityMCP.NL-T">
<system-out><![CDATA[
Applied 3 non-overlapping edits in one atomic call:
- HasTarget(): added "// validated access"
- ApplyBlend(): added "// safe animation"
- End-of-class: added "// end of test modifications"
validate_script: OK
]]></system-out>
</testcase>
```
```xml
<testcase name="T-G — Path Normalization Test" classname="UnityMCP.NL-T">
<system-out><![CDATA[
Edit via unity://path/... succeeded.
Same edit via Assets/... returned stale_file, retried with updated hash: OK.
]]></system-out>
</testcase>
```
```xml
<testcase name="T-H — Validation on Modified File" classname="UnityMCP.NL-T">
<system-out><![CDATA[
validate_script(level:"standard"): OK on the modified file.
]]></system-out>
</testcase>
```
```xml
<testcase name="T-I — Failure Surface Testing" classname="UnityMCP.NL-T">
<system-out><![CDATA[
Overlapping edit: failed cleanly (error captured).
Stale hash edit: failed cleanly (error captured).
File unchanged.
]]></system-out>
</testcase>
```
```xml
<testcase name="T-J — Idempotency on Modified File" classname="UnityMCP.NL-T">
<system-out><![CDATA[
Insert marker after "// Tail test C": OK.
Insert same marker again: no_op: true.
regex_remove marker: OK.
regex_remove again: no_op: true.
validate_script: OK.
]]></system-out>
</testcase>

18
.claude/settings.json Normal file
View File

@ -0,0 +1,18 @@
{
"permissions": {
"allow": [
"mcp__unity",
"Edit(reports/**)",
"MultiEdit(reports/**)"
],
"deny": [
"Bash",
"WebFetch",
"WebSearch",
"Task",
"TodoWrite",
"NotebookEdit",
"NotebookRead"
]
}
}

356
.github/workflows/claude-nl-suite-mini.yml vendored
View File

@ -1,356 +0,0 @@
name: Claude Mini NL Test Suite (Unity live)
on:
workflow_dispatch: {}
permissions:
contents: read
checks: write
concurrency:
group: ${{ github.workflow }}-${{ github.ref }}
cancel-in-progress: true
env:
UNITY_VERSION: 2021.3.45f1
UNITY_IMAGE: unityci/editor:ubuntu-2021.3.45f1-linux-il2cpp-3
UNITY_CACHE_ROOT: /home/runner/work/_temp/_github_home
jobs:
nl-suite:
if: github.event_name == 'workflow_dispatch'
runs-on: ubuntu-latest
timeout-minutes: 60
env:
JUNIT_OUT: reports/junit-nl-suite.xml
MD_OUT: reports/junit-nl-suite.md
steps:
# ---------- Detect secrets ----------
- name: Detect secrets (outputs)
id: detect
env:
UNITY_LICENSE: ${{ secrets.UNITY_LICENSE }}
UNITY_EMAIL: ${{ secrets.UNITY_EMAIL }}
UNITY_PASSWORD: ${{ secrets.UNITY_PASSWORD }}
UNITY_SERIAL: ${{ secrets.UNITY_SERIAL }}
ANTHROPIC_API_KEY: ${{ secrets.ANTHROPIC_API_KEY }}
run: |
set -e
if [ -n "$ANTHROPIC_API_KEY" ]; then echo "anthropic_ok=true" >> "$GITHUB_OUTPUT"; else echo "anthropic_ok=false" >> "$GITHUB_OUTPUT"; fi
if [ -n "$UNITY_LICENSE" ] || { [ -n "$UNITY_EMAIL" ] && [ -n "$UNITY_PASSWORD" ]; } || [ -n "$UNITY_SERIAL" ]; then
echo "unity_ok=true" >> "$GITHUB_OUTPUT"
else
echo "unity_ok=false" >> "$GITHUB_OUTPUT"
fi
- uses: actions/checkout@v4
with:
fetch-depth: 0
# ---------- Python env for MCP server (uv) ----------
- uses: astral-sh/setup-uv@v4
with:
python-version: '3.11'
- name: Install MCP server
run: |
set -eux
uv venv
echo "VIRTUAL_ENV=$GITHUB_WORKSPACE/.venv" >> "$GITHUB_ENV"
echo "$GITHUB_WORKSPACE/.venv/bin" >> "$GITHUB_PATH"
if [ -f UnityMcpBridge/UnityMcpServer~/src/pyproject.toml ]; then
uv pip install -e UnityMcpBridge/UnityMcpServer~/src
elif [ -f UnityMcpBridge/UnityMcpServer~/src/requirements.txt ]; then
uv pip install -r UnityMcpBridge/UnityMcpServer~/src/requirements.txt
elif [ -f UnityMcpBridge/UnityMcpServer~/pyproject.toml ]; then
uv pip install -e UnityMcpBridge/UnityMcpServer~/
elif [ -f UnityMcpBridge/UnityMcpServer~/requirements.txt ]; then
uv pip install -r UnityMcpBridge/UnityMcpServer~/requirements.txt
else
echo "No MCP Python deps found (skipping)"
fi
# ---------- License prime on host (handles ULF or EBL) ----------
- name: Prime Unity license on host (GameCI)
if: steps.detect.outputs.unity_ok == 'true'
uses: game-ci/unity-test-runner@v4
env:
UNITY_LICENSE: ${{ secrets.UNITY_LICENSE }}
UNITY_EMAIL: ${{ secrets.UNITY_EMAIL }}
UNITY_PASSWORD: ${{ secrets.UNITY_PASSWORD }}
UNITY_SERIAL: ${{ secrets.UNITY_SERIAL }}
with:
projectPath: TestProjects/UnityMCPTests
testMode: EditMode
customParameters: -runTests -testFilter __NoSuchTest__ -batchmode -nographics
unityVersion: ${{ env.UNITY_VERSION }}
# (Optional) Show where the license actually got written
- name: Inspect GameCI license caches (host)
if: steps.detect.outputs.unity_ok == 'true'
run: |
set -eux
find "${{ env.UNITY_CACHE_ROOT }}" -maxdepth 4 \( -path "*/.cache" -prune -o -type f \( -name '*.ulf' -o -name 'user.json' \) -print \) 2>/dev/null || true
# ---------- Clean any stale MCP status from previous runs ----------
- name: Clean old MCP status
run: |
set -eux
mkdir -p "$HOME/.unity-mcp"
rm -f "$HOME/.unity-mcp"/unity-mcp-status-*.json || true
# ---------- Start headless Unity that stays up (bridge enabled) ----------
- name: Start Unity (persistent bridge)
if: steps.detect.outputs.unity_ok == 'true'
env:
UNITY_EMAIL: ${{ secrets.UNITY_EMAIL }}
UNITY_PASSWORD: ${{ secrets.UNITY_PASSWORD }}
UNITY_SERIAL: ${{ secrets.UNITY_SERIAL }}
run: |
set -eu
if [ ! -d "${{ github.workspace }}/TestProjects/UnityMCPTests/ProjectSettings" ]; then
echo "Unity project not found; failing fast."
exit 1
fi
mkdir -p "$HOME/.unity-mcp"
MANUAL_ARG=()
if [ -f "${UNITY_CACHE_ROOT}/.local/share/unity3d/Unity_lic.ulf" ]; then
MANUAL_ARG=(-manualLicenseFile /root/.local/share/unity3d/Unity_lic.ulf)
fi
EBL_ARGS=()
[ -n "${UNITY_SERIAL:-}" ] && EBL_ARGS+=(-serial "$UNITY_SERIAL")
[ -n "${UNITY_EMAIL:-}" ] && EBL_ARGS+=(-username "$UNITY_EMAIL")
[ -n "${UNITY_PASSWORD:-}" ] && EBL_ARGS+=(-password "$UNITY_PASSWORD")
docker rm -f unity-mcp >/dev/null 2>&1 || true
docker run -d --name unity-mcp --network host \
-e HOME=/root \
-e UNITY_MCP_ALLOW_BATCH=1 -e UNITY_MCP_STATUS_DIR=/root/.unity-mcp \
-e UNITY_MCP_BIND_HOST=127.0.0.1 \
-v "${{ github.workspace }}:/workspace" -w /workspace \
-v "${{ env.UNITY_CACHE_ROOT }}:/root" \
-v "$HOME/.unity-mcp:/root/.unity-mcp" \
${{ env.UNITY_IMAGE }} /opt/unity/Editor/Unity -batchmode -nographics -logFile - \
-stackTraceLogType Full \
-projectPath /workspace/TestProjects/UnityMCPTests \
"${MANUAL_ARG[@]}" \
"${EBL_ARGS[@]}" \
-executeMethod MCPForUnity.Editor.MCPForUnityBridge.StartAutoConnect
# ---------- Wait for Unity bridge (fail fast if not running/ready) ----------
- name: Wait for Unity bridge (robust)
if: steps.detect.outputs.unity_ok == 'true'
run: |
set -euo pipefail
if ! docker ps --format '{{.Names}}' | grep -qx 'unity-mcp'; then
echo "Unity container failed to start"; docker ps -a || true; exit 1
fi
docker logs -f unity-mcp 2>&1 | sed -E 's/((serial|license|password|token)[^[:space:]]*)/[REDACTED]/ig' & LOGPID=$!
deadline=$((SECONDS+420)); READY=0
try_connect_host() {
P="$1"
timeout 1 bash -lc "exec 3<>/dev/tcp/127.0.0.1/$P; head -c 8 <&3 >/dev/null" && return 0 || true
if command -v nc >/dev/null 2>&1; then nc -6 -z ::1 "$P" && return 0 || true; fi
return 1
}
# in-container probe will try IPv4 then IPv6 via nc or /dev/tcp
while [ $SECONDS -lt $deadline ]; do
if docker logs unity-mcp 2>&1 | grep -qE "MCP Bridge listening|Bridge ready|Server started"; then
READY=1; echo "Bridge ready (log markers)"; break
fi
PORT=$(python -c "import os,glob,json,sys,time; b=os.path.expanduser('~/.unity-mcp'); fs=sorted(glob.glob(os.path.join(b,'unity-mcp-status-*.json')), key=os.path.getmtime, reverse=True); print(next((json.load(open(f,'r',encoding='utf-8')).get('unity_port') for f in fs if time.time()-os.path.getmtime(f)<=300 and json.load(open(f,'r',encoding='utf-8')).get('unity_port')), '' ))" 2>/dev/null || true)
if [ -n "${PORT:-}" ] && { try_connect_host "$PORT" || docker exec unity-mcp bash -lc "timeout 1 bash -lc 'exec 3<>/dev/tcp/127.0.0.1/$PORT' || (command -v nc >/dev/null 2>&1 && nc -6 -z ::1 $PORT)"; }; then
READY=1; echo "Bridge ready on port $PORT"; break
fi
if docker logs unity-mcp 2>&1 | grep -qE "No valid Unity Editor license|Token not found in cache|com\.unity\.editor\.headless"; then
echo "Licensing error detected"; break
fi
sleep 2
done
kill $LOGPID || true
if [ "$READY" != "1" ]; then
echo "Bridge not ready; diagnostics:"
echo "== status files =="; ls -la "$HOME/.unity-mcp" || true
echo "== status contents =="; for f in "$HOME"/.unity-mcp/unity-mcp-status-*.json; do [ -f "$f" ] && { echo "--- $f"; sed -n '1,120p' "$f"; }; done
echo "== sockets (inside container) =="; docker exec unity-mcp bash -lc 'ss -lntp || netstat -tulpen || true'
echo "== tail of Unity log =="
docker logs --tail 200 unity-mcp | sed -E 's/((serial|license|password|token)[^[:space:]]*)/[REDACTED]/ig' || true
exit 1
fi
# ---------- Make MCP config available to the action ----------
- name: Write MCP config (.claude/mcp.json)
run: |
set -eux
mkdir -p .claude
cat > .claude/mcp.json <<JSON
{
"mcpServers": {
"unity": {
"command": "uv",
"args": ["run","--active","--directory","UnityMcpBridge/UnityMcpServer~/src","python","server.py"],
"transport": { "type": "stdio" },
"env": {
"PYTHONUNBUFFERED": "1",
"MCP_LOG_LEVEL": "debug",
"UNITY_PROJECT_ROOT": "$GITHUB_WORKSPACE/TestProjects/UnityMCPTests"
}
}
}
}
JSON
# ---------- Ensure reports dir exists ----------
- name: Prepare reports
run: |
set -eux
mkdir -p reports
# ---------- Run full NL suite once ----------
- name: Run Claude NL suite (single pass)
uses: anthropics/claude-code-base-action@beta
if: steps.detect.outputs.anthropic_ok == 'true' && steps.detect.outputs.unity_ok == 'true'
env:
JUNIT_OUT: reports/junit-nl-suite.xml
MD_OUT: reports/junit-nl-suite.md
with:
use_node_cache: false
prompt_file: .claude/prompts/nl-unity-claude-tests-mini.md
mcp_config: .claude/mcp.json
allowed_tools: "Write,mcp__unity__manage_editor,mcp__unity__list_resources,mcp__unity__read_resource,mcp__unity__apply_text_edits,mcp__unity__script_apply_edits,mcp__unity__validate_script,mcp__unity__find_in_file, mcp__unity__read_console"
disallowed_tools: "TodoWrite,Task"
model: "claude-3-7-sonnet-latest"
timeout_minutes: "30"
anthropic_api_key: ${{ secrets.ANTHROPIC_API_KEY }}
- name: Normalize JUnit for consumer actions (strong)
if: always()
shell: bash
run: |
python3 - <<'PY'
from pathlib import Path
import xml.etree.ElementTree as ET
import os
def localname(tag: str) -> str:
return tag.rsplit('}', 1)[-1] if '}' in tag else tag
src = Path(os.environ.get('JUNIT_OUT', 'reports/junit-nl-suite.xml'))
out = Path('reports/junit-for-actions.xml')
out.parent.mkdir(parents=True, exist_ok=True)
if not src.exists():
# Try to use any existing XML as a source (e.g., claude-nl-tests.xml)
candidates = sorted(Path('reports').glob('*.xml'))
if candidates:
src = candidates[0]
else:
print("WARN: no XML source found for normalization")
if src.exists():
try:
root = ET.parse(src).getroot()
rtag = localname(root.tag)
if rtag == 'testsuites' and len(root) == 1 and localname(root[0].tag) == 'testsuite':
ET.ElementTree(root[0]).write(out, encoding='utf-8', xml_declaration=True)
else:
out.write_bytes(src.read_bytes())
except Exception as e:
print("Normalization error:", e)
out.write_bytes(src.read_bytes())
# Always create a second copy with a junit-* name so wildcard patterns match too
if out.exists():
Path('reports/junit-nl-suite-copy.xml').write_bytes(out.read_bytes())
PY
- name: "Debug: list report files"
if: always()
shell: bash
run: |
set -eux
ls -la reports || true
shopt -s nullglob
for f in reports/*.xml; do
echo "===== $f ====="
head -n 40 "$f" || true
done
# sanitize only the markdown (does not touch JUnit xml)
- name: Sanitize markdown (all shards)
if: always()
run: |
set -eu
python - <<'PY'
from pathlib import Path
rp=Path('reports')
rp.mkdir(parents=True, exist_ok=True)
for p in rp.glob('*.md'):
b=p.read_bytes().replace(b'\x00', b'')
s=b.decode('utf-8','replace').replace('\r\n','\n')
p.write_text(s, encoding='utf-8', newline='\n')
PY
- name: NL/T details → Job Summary
if: always()
run: |
echo "## Unity NL/T Editing Suite — Full Coverage" >> $GITHUB_STEP_SUMMARY
python - <<'PY' >> $GITHUB_STEP_SUMMARY
from pathlib import Path
p = Path('reports/junit-nl-suite.md') if Path('reports/junit-nl-suite.md').exists() else Path('reports/claude-nl-tests.md')
if p.exists():
text = p.read_bytes().decode('utf-8', 'replace')
MAX = 65000
print(text[:MAX])
if len(text) > MAX:
print("\n\n_…truncated in summary; full report is in artifacts._")
else:
print("_No markdown report found._")
PY
- name: Fallback JUnit if missing
if: always()
run: |
set -eu
mkdir -p reports
if [ ! -f reports/junit-for-actions.xml ]; then
printf '%s\n' \
'<?xml version="1.0" encoding="UTF-8"?>' \
'<testsuite name="UnityMCP.NL-T" tests="1" failures="1" time="0">' \
' <testcase classname="UnityMCP.NL-T" name="NL-Suite.Execution" time="0.0">' \
' <failure><![CDATA[No JUnit was produced by the NL suite step. See the '"'"'Run Claude NL suite (single pass)'"'"' logs.]]></failure>' \
' </testcase>' \
'</testsuite>' \
> reports/junit-for-actions.xml
fi
- name: Publish JUnit reports
if: always()
uses: mikepenz/action-junit-report@v5
with:
report_paths: 'reports/junit-for-actions.xml'
include_passed: true
detailed_summary: true
annotate_notice: true
require_tests: false
fail_on_parse_error: true
- name: Upload artifacts
if: always()
uses: actions/upload-artifact@v4
with:
name: claude-nl-suite-artifacts
path: reports/**
# ---------- Always stop Unity ----------
- name: Stop Unity
if: always()
run: |
docker logs --tail 400 unity-mcp | sed -E 's/((serial|license|password|token)[^[:space:]]*)/[REDACTED]/ig' || true
docker rm -f unity-mcp || true

808
.github/workflows/claude-nl-suite.yml vendored
View File

@ -1,7 +1,6 @@
name: Claude NL/T Full Suite (Unity live) name: Claude NL/T Full Suite (Unity live)
on: on: [workflow_dispatch]
workflow_dispatch: {}
permissions: permissions:
contents: read contents: read
@ -12,13 +11,10 @@ concurrency:
cancel-in-progress: true cancel-in-progress: true
env: env:
UNITY_VERSION: 2021.3.45f1
UNITY_IMAGE: unityci/editor:ubuntu-2021.3.45f1-linux-il2cpp-3 UNITY_IMAGE: unityci/editor:ubuntu-2021.3.45f1-linux-il2cpp-3
UNITY_CACHE_ROOT: /home/runner/work/_temp/_github_home
jobs: jobs:
nl-suite: nl-suite:
if: github.event_name == 'workflow_dispatch'
runs-on: ubuntu-latest runs-on: ubuntu-latest
timeout-minutes: 60 timeout-minutes: 60
env: env:
@ -38,7 +34,7 @@ jobs:
run: | run: |
set -e set -e
if [ -n "$ANTHROPIC_API_KEY" ]; then echo "anthropic_ok=true" >> "$GITHUB_OUTPUT"; else echo "anthropic_ok=false" >> "$GITHUB_OUTPUT"; fi if [ -n "$ANTHROPIC_API_KEY" ]; then echo "anthropic_ok=true" >> "$GITHUB_OUTPUT"; else echo "anthropic_ok=false" >> "$GITHUB_OUTPUT"; fi
if [ -n "$UNITY_LICENSE" ] || { [ -n "$UNITY_EMAIL" ] && [ -n "$UNITY_PASSWORD" ]; } || [ -n "$UNITY_SERIAL" ]; then if [ -n "$UNITY_LICENSE" ] || { [ -n "$UNITY_EMAIL" ] && [ -n "$UNITY_PASSWORD" ]; }; then
echo "unity_ok=true" >> "$GITHUB_OUTPUT" echo "unity_ok=true" >> "$GITHUB_OUTPUT"
else else
echo "unity_ok=false" >> "$GITHUB_OUTPUT" echo "unity_ok=false" >> "$GITHUB_OUTPUT"
@ -71,27 +67,119 @@ jobs:
echo "No MCP Python deps found (skipping)" echo "No MCP Python deps found (skipping)"
fi fi
# ---------- License prime on host (GameCI) ---------- # --- Licensing: allow both ULF and EBL when available ---
- name: Prime Unity license on host (GameCI) - name: Decide license sources
if: steps.detect.outputs.unity_ok == 'true' id: lic
uses: game-ci/unity-test-runner@v4 shell: bash
env: env:
UNITY_LICENSE: ${{ secrets.UNITY_LICENSE }} UNITY_LICENSE: ${{ secrets.UNITY_LICENSE }}
UNITY_EMAIL: ${{ secrets.UNITY_EMAIL }} UNITY_EMAIL: ${{ secrets.UNITY_EMAIL }}
UNITY_PASSWORD: ${{ secrets.UNITY_PASSWORD }} UNITY_PASSWORD: ${{ secrets.UNITY_PASSWORD }}
UNITY_SERIAL: ${{ secrets.UNITY_SERIAL }} UNITY_SERIAL: ${{ secrets.UNITY_SERIAL }}
with:
projectPath: TestProjects/UnityMCPTests
testMode: EditMode
customParameters: -runTests -testFilter __NoSuchTest__ -batchmode -nographics
unityVersion: ${{ env.UNITY_VERSION }}
# (Optional) Inspect license caches
- name: Inspect GameCI license caches (host)
if: steps.detect.outputs.unity_ok == 'true'
run: | run: |
set -eux set -eu
find "${{ env.UNITY_CACHE_ROOT }}" -maxdepth 4 \( -path "*/.cache" -prune -o -type f \( -name '*.ulf' -o -name 'user.json' \) -print \) 2>/dev/null || true use_ulf=false; use_ebl=false
[[ -n "${UNITY_LICENSE:-}" ]] && use_ulf=true
[[ -n "${UNITY_EMAIL:-}" && -n "${UNITY_PASSWORD:-}" ]] && use_ebl=true
echo "use_ulf=$use_ulf" >> "$GITHUB_OUTPUT"
echo "use_ebl=$use_ebl" >> "$GITHUB_OUTPUT"
echo "has_serial=$([[ -n "${UNITY_SERIAL:-}" ]] && echo true || echo false)" >> "$GITHUB_OUTPUT"
- name: Stage Unity .ulf license (from secret)
if: steps.lic.outputs.use_ulf == 'true'
id: ulf
env:
UNITY_LICENSE: ${{ secrets.UNITY_LICENSE }}
shell: bash
run: |
set -eu
mkdir -p "$RUNNER_TEMP/unity-license-ulf" "$RUNNER_TEMP/unity-local/Unity"
f="$RUNNER_TEMP/unity-license-ulf/Unity_lic.ulf"
if printf "%s" "$UNITY_LICENSE" | base64 -d - >/dev/null 2>&1; then
printf "%s" "$UNITY_LICENSE" | base64 -d - > "$f"
else
printf "%s" "$UNITY_LICENSE" > "$f"
fi
chmod 600 "$f" || true
# If someone pasted an entitlement XML into UNITY_LICENSE by mistake, re-home it:
if head -c 100 "$f" | grep -qi '<\?xml'; then
mkdir -p "$RUNNER_TEMP/unity-config/Unity/licenses"
mv "$f" "$RUNNER_TEMP/unity-config/Unity/licenses/UnityEntitlementLicense.xml"
echo "ok=false" >> "$GITHUB_OUTPUT"
elif grep -qi '<Signature>' "$f"; then
# provide it in the standard local-share path too
cp -f "$f" "$RUNNER_TEMP/unity-local/Unity/Unity_lic.ulf"
echo "ok=true" >> "$GITHUB_OUTPUT"
else
echo "ok=false" >> "$GITHUB_OUTPUT"
fi
# --- Activate via EBL inside the same Unity image (writes host-side entitlement) ---
- name: Activate Unity (EBL via container - host-mount)
if: steps.lic.outputs.use_ebl == 'true'
shell: bash
env:
UNITY_IMAGE: ${{ env.UNITY_IMAGE }}
UNITY_EMAIL: ${{ secrets.UNITY_EMAIL }}
UNITY_PASSWORD: ${{ secrets.UNITY_PASSWORD }}
UNITY_SERIAL: ${{ secrets.UNITY_SERIAL }}
run: |
set -euxo pipefail
# host dirs to receive the full Unity config and local-share
mkdir -p "$RUNNER_TEMP/unity-config" "$RUNNER_TEMP/unity-local"
# Try Pro first if serial is present, otherwise named-user EBL.
docker run --rm --network host \
-e HOME=/root \
-e UNITY_EMAIL -e UNITY_PASSWORD -e UNITY_SERIAL \
-v "$RUNNER_TEMP/unity-config:/root/.config/unity3d" \
-v "$RUNNER_TEMP/unity-local:/root/.local/share/unity3d" \
"$UNITY_IMAGE" bash -lc '
set -euxo pipefail
if [[ -n "${UNITY_SERIAL:-}" ]]; then
/opt/unity/Editor/Unity -batchmode -nographics -logFile - \
-username "$UNITY_EMAIL" -password "$UNITY_PASSWORD" -serial "$UNITY_SERIAL" -quit || true
else
/opt/unity/Editor/Unity -batchmode -nographics -logFile - \
-username "$UNITY_EMAIL" -password "$UNITY_PASSWORD" -quit || true
fi
ls -la /root/.config/unity3d/Unity/licenses || true
'
# Verify entitlement written to host mount; allow ULF-only runs to proceed
if ! find "$RUNNER_TEMP/unity-config" -type f -iname "*.xml" | grep -q .; then
if [[ "${{ steps.ulf.outputs.ok }}" == "true" ]]; then
echo "EBL entitlement not found; proceeding with ULF-only (ok=true)."
else
echo "No entitlement produced and no valid ULF; cannot continue." >&2
exit 1
fi
fi
# EBL entitlement is already written directly to $RUNNER_TEMP/unity-config by the activation step
# ---------- Warm up project (import Library once) ----------
- name: Warm up project (import Library once)
if: steps.lic.outputs.use_ulf == 'true' || steps.lic.outputs.use_ebl == 'true'
shell: bash
env:
UNITY_IMAGE: ${{ env.UNITY_IMAGE }}
ULF_OK: ${{ steps.ulf.outputs.ok }}
run: |
set -euxo pipefail
manual_args=()
if [[ "${ULF_OK:-false}" == "true" ]]; then
manual_args=(-manualLicenseFile "/root/.local/share/unity3d/Unity/Unity_lic.ulf")
fi
docker run --rm --network host \
-e HOME=/root \
-v "${{ github.workspace }}:/workspace" -w /workspace \
-v "$RUNNER_TEMP/unity-config:/root/.config/unity3d" \
-v "$RUNNER_TEMP/unity-local:/root/.local/share/unity3d" \
"$UNITY_IMAGE" /opt/unity/Editor/Unity -batchmode -nographics -logFile - \
-projectPath /workspace/TestProjects/UnityMCPTests \
"${manual_args[@]}" \
-quit
# ---------- Clean old MCP status ---------- # ---------- Clean old MCP status ----------
- name: Clean old MCP status - name: Clean old MCP status
@ -102,81 +190,91 @@ jobs:
# ---------- Start headless Unity (persistent bridge) ---------- # ---------- Start headless Unity (persistent bridge) ----------
- name: Start Unity (persistent bridge) - name: Start Unity (persistent bridge)
if: steps.detect.outputs.unity_ok == 'true' if: steps.lic.outputs.use_ulf == 'true' || steps.lic.outputs.use_ebl == 'true'
shell: bash
env: env:
UNITY_EMAIL: ${{ secrets.UNITY_EMAIL }} UNITY_IMAGE: ${{ env.UNITY_IMAGE }}
UNITY_PASSWORD: ${{ secrets.UNITY_PASSWORD }} ULF_OK: ${{ steps.ulf.outputs.ok }}
UNITY_SERIAL: ${{ secrets.UNITY_SERIAL }}
run: | run: |
set -eu set -euxo pipefail
if [ ! -d "${{ github.workspace }}/TestProjects/UnityMCPTests/ProjectSettings" ]; then manual_args=()
echo "Unity project not found; failing fast." if [[ "${ULF_OK:-false}" == "true" ]]; then
exit 1 manual_args=(-manualLicenseFile "/root/.local/share/unity3d/Unity/Unity_lic.ulf")
fi fi
mkdir -p "$HOME/.unity-mcp"
MANUAL_ARG=() mkdir -p "$RUNNER_TEMP/unity-status"
if [ -f "${UNITY_CACHE_ROOT}/.local/share/unity3d/Unity_lic.ulf" ]; then
MANUAL_ARG=(-manualLicenseFile /root/.local/share/unity3d/Unity_lic.ulf)
fi
EBL_ARGS=()
[ -n "${UNITY_SERIAL:-}" ] && EBL_ARGS+=(-serial "$UNITY_SERIAL")
[ -n "${UNITY_EMAIL:-}" ] && EBL_ARGS+=(-username "$UNITY_EMAIL")
[ -n "${UNITY_PASSWORD:-}" ] && EBL_ARGS+=(-password "$UNITY_PASSWORD")
docker rm -f unity-mcp >/dev/null 2>&1 || true docker rm -f unity-mcp >/dev/null 2>&1 || true
docker run -d --name unity-mcp --network host \ docker run -d --name unity-mcp --network host \
-e HOME=/root \ -e HOME=/root \
-e UNITY_MCP_ALLOW_BATCH=1 -e UNITY_MCP_STATUS_DIR=/root/.unity-mcp \ -e UNITY_MCP_ALLOW_BATCH=1 \
-e UNITY_MCP_STATUS_DIR=/root/.unity-mcp \
-e UNITY_MCP_BIND_HOST=127.0.0.1 \ -e UNITY_MCP_BIND_HOST=127.0.0.1 \
-v "${{ github.workspace }}:/workspace" -w /workspace \ -v "${{ github.workspace }}:/workspace" -w /workspace \
-v "${{ env.UNITY_CACHE_ROOT }}:/root" \ -v "$RUNNER_TEMP/unity-status:/root/.unity-mcp" \
-v "$HOME/.unity-mcp:/root/.unity-mcp" \ -v "$RUNNER_TEMP/unity-config:/root/.config/unity3d:ro" \
${{ env.UNITY_IMAGE }} /opt/unity/Editor/Unity -batchmode -nographics -logFile - \ -v "$RUNNER_TEMP/unity-local:/root/.local/share/unity3d:ro" \
"$UNITY_IMAGE" /opt/unity/Editor/Unity -batchmode -nographics -logFile - \
-stackTraceLogType Full \ -stackTraceLogType Full \
-projectPath /workspace/TestProjects/UnityMCPTests \ -projectPath /workspace/TestProjects/UnityMCPTests \
"${MANUAL_ARG[@]}" \ "${manual_args[@]}" \
"${EBL_ARGS[@]}" \
-executeMethod MCPForUnity.Editor.MCPForUnityBridge.StartAutoConnect -executeMethod MCPForUnity.Editor.MCPForUnityBridge.StartAutoConnect
# ---------- Wait for Unity bridge ---------- # ---------- Wait for Unity bridge ----------
- name: Wait for Unity bridge (robust) - name: Wait for Unity bridge (robust)
if: steps.detect.outputs.unity_ok == 'true' shell: bash
run: | run: |
set -euo pipefail set -euo pipefail
if ! docker ps --format '{{.Names}}' | grep -qx 'unity-mcp'; then deadline=$((SECONDS+900)) # 15 min max
echo "Unity container failed to start"; docker ps -a || true; exit 1 fatal_after=$((SECONDS+120)) # give licensing 2 min to settle
fi
docker logs -f unity-mcp 2>&1 | sed -E 's/((serial|license|password|token)[^[:space:]]*)/[REDACTED]/ig' & LOGPID=$! # Fail fast only if container actually died
deadline=$((SECONDS+420)); READY=0 st="$(docker inspect -f '{{.State.Status}} {{.State.ExitCode}}' unity-mcp 2>/dev/null || true)"
try_connect_host() { case "$st" in exited*|dead*) docker logs unity-mcp --tail 200 | sed -E 's/((email|serial|license|password|token)[^[:space:]]*)/[REDACTED]/Ig'; exit 1;; esac
P="$1"
timeout 1 bash -lc "exec 3<>/dev/tcp/127.0.0.1/$P; head -c 8 <&3 >/dev/null" && return 0 || true # Patterns
if command -v nc >/dev/null 2>&1; then nc -6 -z ::1 "$P" && return 0 || true; fi ok_pat='(Bridge|MCP(For)?Unity|AutoConnect).*(listening|ready|started|port|bound)'
return 1 # Only truly fatal signals; allow transient "Licensing::..." chatter
} license_fatal='No valid Unity|License is not active|cannot load ULF|Signature element not found|Token not found|0 entitlement|Entitlement.*(failed|denied)|License (activation|return|renewal).*(failed|expired|denied)'
while [ $SECONDS -lt $deadline ]; do while [ $SECONDS -lt $deadline ]; do
if docker logs unity-mcp 2>&1 | grep -qE "MCP Bridge listening|Bridge ready|Server started"; then logs="$(docker logs unity-mcp 2>&1 || true)"
READY=1; echo "Bridge ready (log markers)"; break
# 1) Primary: status JSON exposes TCP port
port="$(jq -r '.unity_port // empty' "$RUNNER_TEMP"/unity-status/unity-mcp-status-*.json 2>/dev/null | head -n1 || true)"
if [[ -n "${port:-}" ]] && timeout 1 bash -lc "exec 3<>/dev/tcp/127.0.0.1/$port"; then
echo "Bridge ready on port $port"
exit 0
fi fi
PORT=$(python3 -c "import os,glob,json,sys,time; b=os.path.expanduser('~/.unity-mcp'); fs=sorted(glob.glob(os.path.join(b,'unity-mcp-status-*.json')), key=os.path.getmtime, reverse=True); print(next((json.load(open(f,'r',encoding='utf-8')).get('unity_port') for f in fs if time.time()-os.path.getmtime(f)<=300 and json.load(open(f,'r',encoding='utf-8')).get('unity_port')), '' ))" 2>/dev/null || true)
if [ -n "${PORT:-}" ] && { try_connect_host "$PORT" || docker exec unity-mcp bash -lc "timeout 1 bash -lc 'exec 3<>/dev/tcp/127.0.0.1/$PORT' || (command -v nc >/dev/null 2>&1 && nc -6 -z ::1 $PORT)"; }; then # 2) Secondary: log markers
READY=1; echo "Bridge ready on port $PORT"; break if echo "$logs" | grep -qiE "$ok_pat"; then
echo "Bridge ready (log markers)"
exit 0
fi fi
if docker logs unity-mcp 2>&1 | grep -qE "No valid Unity Editor license|Token not found in cache|com\.unity\.editor\.headless"; then
echo "Licensing error detected"; break # Only treat license failures as fatal *after* warm-up
fi if [ $SECONDS -ge $fatal_after ] && echo "$logs" | grep -qiE "$license_fatal"; then
sleep 2 echo "::error::Fatal licensing signal detected after warm-up"
done echo "$logs" | tail -n 200 | sed -E 's/((email|serial|license|password|token)[^[:space:]]*)/[REDACTED]/Ig'
kill $LOGPID || true
if [ "$READY" != "1" ]; then
echo "Bridge not ready; diagnostics:"
echo "== status files =="; ls -la "$HOME/.unity-mcp" || true
echo "== status contents =="; for f in "$HOME"/.unity-mcp/unity-mcp-status-*.json; do [ -f "$f" ] && { echo "--- $f"; sed -n '1,120p' "$f"; }; done
echo "== sockets (inside container) =="; docker exec unity-mcp bash -lc 'ss -lntp || netstat -tulpen || true'
echo "== tail of Unity log =="
docker logs --tail 200 unity-mcp | sed -E 's/((serial|license|password|token)[^[:space:]]*)/[REDACTED]/ig' || true
exit 1 exit 1
fi fi
# If the container dies mid-wait, bail
st="$(docker inspect -f '{{.State.Status}}' unity-mcp 2>/dev/null || true)"
if [[ "$st" != "running" ]]; then
echo "::error::Unity container exited during wait"; docker logs unity-mcp --tail 200 | sed -E 's/((email|serial|license|password|token)[^[:space:]]*)/[REDACTED]/Ig'
exit 1
fi
sleep 2
done
echo "::error::Bridge not ready before deadline"
docker logs unity-mcp --tail 200 | sed -E 's/((email|serial|license|password|token)[^[:space:]]*)/[REDACTED]/Ig'
exit 1
# (moved) — return license after Unity is stopped
# ---------- MCP client config ---------- # ---------- MCP client config ----------
- name: Write MCP config (.claude/mcp.json) - name: Write MCP config (.claude/mcp.json)
run: | run: |
@ -192,19 +290,46 @@ jobs:
"env": { "env": {
"PYTHONUNBUFFERED": "1", "PYTHONUNBUFFERED": "1",
"MCP_LOG_LEVEL": "debug", "MCP_LOG_LEVEL": "debug",
"UNITY_PROJECT_ROOT": "$GITHUB_WORKSPACE/TestProjects/UnityMCPTests" "UNITY_PROJECT_ROOT": "$GITHUB_WORKSPACE/TestProjects/UnityMCPTests",
"UNITY_MCP_STATUS_DIR": "$RUNNER_TEMP/unity-status",
"UNITY_MCP_HOST": "127.0.0.1"
} }
} }
} }
} }
JSON JSON
- name: Pin Claude tool permissions (.claude/settings.json)
run: |
set -eux
mkdir -p .claude
cat > .claude/settings.json <<'JSON'
{
"permissions": {
"allow": [
"mcp__unity",
"Edit(reports/**)"
],
"deny": [
"Bash",
"MultiEdit",
"WebFetch",
"WebSearch",
"Task",
"TodoWrite",
"NotebookEdit",
"NotebookRead"
]
}
}
JSON
# ---------- Reports & helper ---------- # ---------- Reports & helper ----------
- name: Prepare reports and dirs - name: Prepare reports and dirs
run: | run: |
set -eux set -eux
rm -f reports/*.xml reports/*.md || true rm -f reports/*.xml reports/*.md || true
mkdir -p reports reports/_snapshots scripts mkdir -p reports reports/_snapshots reports/_staging
- name: Create report skeletons - name: Create report skeletons
run: | run: |
@ -219,79 +344,299 @@ jobs:
XML XML
printf '# Unity NL/T Editing Suite Test Results\n\n' > "$MD_OUT" printf '# Unity NL/T Editing Suite Test Results\n\n' > "$MD_OUT"
- name: Write safe revert helper (scripts/nlt-revert.sh) - name: Verify Unity bridge status/port
shell: bash
run: | run: |
set -eux set -euxo pipefail
cat > scripts/nlt-revert.sh <<'BASH' ls -la "$RUNNER_TEMP/unity-status" || true
#!/usr/bin/env bash jq -r . "$RUNNER_TEMP"/unity-status/unity-mcp-status-*.json | sed -n '1,80p' || true
set -euo pipefail
sub="${1:-}"; target_rel="${2:-}"; snap="${3:-}" shopt -s nullglob
WS="${GITHUB_WORKSPACE:-$PWD}" status_files=("$RUNNER_TEMP"/unity-status/unity-mcp-status-*.json)
ROOT="$WS/TestProjects/UnityMCPTests" if ((${#status_files[@]})); then
t_abs="$(realpath -m "$WS/$target_rel")" port="$(grep -hEo '"unity_port"[[:space:]]*:[[:space:]]*[0-9]+' "${status_files[@]}" \
s_abs="$(realpath -m "$WS/$snap")" | sed -E 's/.*: *([0-9]+).*/\1/' | head -n1 || true)"
if [[ "$t_abs" != "$ROOT/Assets/"* ]]; then else
echo "refuse: target outside allowed scope: $t_abs" >&2; exit 2 port=""
fi fi
mkdir -p "$(dirname "$s_abs")"
case "$sub" in
snapshot)
cp -f "$t_abs" "$s_abs"
sha=$(sha256sum "$s_abs" | awk '{print $1}')
echo "snapshot_sha=$sha"
;;
restore)
if [[ ! -f "$s_abs" ]]; then echo "snapshot missing: $s_abs" >&2; exit 3; fi
cp -f "$s_abs" "$t_abs"
touch "$t_abs"
sha=$(sha256sum "$t_abs" | awk '{print $1}')
echo "restored_sha=$sha"
;;
*)
echo "usage: $0 snapshot|restore <target_rel_path> <snapshot_path>" >&2; exit 1
;;
esac
BASH
chmod +x scripts/nlt-revert.sh
# ---------- Snapshot baseline (pre-agent) ---------- echo "unity_port=$port"
- name: Snapshot baseline (pre-agent) if [[ -n "$port" ]]; then
if: steps.detect.outputs.anthropic_ok == 'true' && steps.detect.outputs.unity_ok == 'true' timeout 1 bash -lc "exec 3<>/dev/tcp/127.0.0.1/$port" && echo "TCP OK"
shell: bash fi
run: |
set -euo pipefail # (removed) Revert helper and baseline snapshot are no longer used
TARGET="TestProjects/UnityMCPTests/Assets/Scripts/LongUnityScriptClaudeTest.cs"
SNAP="reports/_snapshots/LongUnityScriptClaudeTest.cs.baseline"
scripts/nlt-revert.sh snapshot "$TARGET" "$SNAP"
# ---------- Run suite ---------- # ---------- Run suite in two passes ----------
- name: Run Claude NL suite (single pass) - name: Run Claude NL pass
uses: anthropics/claude-code-base-action@beta uses: anthropics/claude-code-base-action@beta
if: steps.detect.outputs.anthropic_ok == 'true' && steps.detect.outputs.unity_ok == 'true' if: steps.detect.outputs.anthropic_ok == 'true' && steps.detect.outputs.unity_ok == 'true'
continue-on-error: true continue-on-error: true
with: with:
use_node_cache: false use_node_cache: false
prompt_file: .claude/prompts/nl-unity-suite-full-additive.md prompt_file: .claude/prompts/nl-unity-suite-nl.md
mcp_config: .claude/mcp.json mcp_config: .claude/mcp.json
allowed_tools: >- settings: .claude/settings.json
Write, allowed_tools: "mcp__unity,Edit(reports/**),MultiEdit(reports/**)"
Bash(scripts/nlt-revert.sh:*), disallowed_tools: "Bash,WebFetch,WebSearch,Task,TodoWrite,NotebookEdit,NotebookRead"
mcp__unity__manage_editor, model: claude-3-7-sonnet-20250219
mcp__unity__list_resources, append_system_prompt: |
mcp__unity__read_resource, You are running the NL pass only.
mcp__unity__apply_text_edits, - Emit exactly NL-0, NL-1, NL-2, NL-3, NL-4.
mcp__unity__script_apply_edits, - Write each to reports/${ID}_results.xml.
mcp__unity__validate_script, - Prefer a single MultiEdit(reports/**) batch. Do not emit any T-* tests.
mcp__unity__find_in_file, - Stop after NL-4_results.xml is written.
mcp__unity__read_console,
mcp__unity__get_sha
disallowed_tools: TodoWrite,Task
model: claude-3-7-sonnet-latest
timeout_minutes: "30" timeout_minutes: "30"
anthropic_api_key: ${{ secrets.ANTHROPIC_API_KEY }} anthropic_api_key: ${{ secrets.ANTHROPIC_API_KEY }}
- name: Run Claude T pass A-J
uses: anthropics/claude-code-base-action@beta
if: steps.detect.outputs.anthropic_ok == 'true' && steps.detect.outputs.unity_ok == 'true'
continue-on-error: true
with:
use_node_cache: false
prompt_file: .claude/prompts/nl-unity-suite-t.md
mcp_config: .claude/mcp.json
settings: .claude/settings.json
allowed_tools: "mcp__unity,Edit(reports/**),MultiEdit(reports/**)"
disallowed_tools: "Bash,WebFetch,WebSearch,Task,TodoWrite,NotebookEdit,NotebookRead"
model: claude-3-5-haiku-20241022
append_system_prompt: |
You are running the T pass (A–J) only.
Output requirements:
- Emit exactly 10 test fragments: T-A, T-B, T-C, T-D, T-E, T-F, T-G, T-H, T-I, T-J.
- Write each fragment to reports/${ID}_results.xml (e.g., T-A_results.xml).
- Prefer a single MultiEdit(reports/**) call that writes all ten files in one batch.
- If MultiEdit is not used, emit individual writes for any missing IDs until all ten exist.
- Do not emit any NL-* fragments.
Stop condition:
- After T-J_results.xml is written, stop.
timeout_minutes: "30"
anthropic_api_key: ${{ secrets.ANTHROPIC_API_KEY }}
# (moved) Assert T coverage after staged fragments are promoted
- name: Check T coverage incomplete (pre-retry)
id: t_cov
if: always()
shell: bash
run: |
set -euo pipefail
missing=()
for id in T-A T-B T-C T-D T-E T-F T-G T-H T-I T-J; do
if [[ ! -s "reports/${id}_results.xml" && ! -s "reports/_staging/${id}_results.xml" ]]; then
missing+=("$id")
fi
done
echo "missing=${#missing[@]}" >> "$GITHUB_OUTPUT"
if (( ${#missing[@]} )); then
echo "list=${missing[*]}" >> "$GITHUB_OUTPUT"
fi
- name: Retry T pass (Sonnet) if incomplete
if: steps.t_cov.outputs.missing != '0'
uses: anthropics/claude-code-base-action@beta
with:
use_node_cache: false
prompt_file: .claude/prompts/nl-unity-suite-t.md
mcp_config: .claude/mcp.json
settings: .claude/settings.json
allowed_tools: "mcp__unity,Edit(reports/**),MultiEdit(reports/**)"
disallowed_tools: "Bash,MultiEdit(/!(reports/**)),WebFetch,WebSearch,Task,TodoWrite,NotebookEdit,NotebookRead"
model: claude-3-7-sonnet-20250219
fallback_model: claude-3-5-haiku-20241022
append_system_prompt: |
You are running the T pass only.
Output requirements:
- Emit exactly 10 test fragments: T-A, T-B, T-C, T-D, T-E, T-F, T-G, T-H, T-I, T-J.
- Write each fragment to reports/${ID}_results.xml (e.g., T-A_results.xml).
- Prefer a single MultiEdit(reports/**) call that writes all ten files in one batch.
- If MultiEdit is not used, emit individual writes for any missing IDs until all ten exist.
- Do not emit any NL-* fragments.
Stop condition:
- After T-J_results.xml is written, stop.
timeout_minutes: "30"
anthropic_api_key: ${{ secrets.ANTHROPIC_API_KEY }}
- name: Re-assert T coverage (post-retry)
if: always()
shell: bash
run: |
set -euo pipefail
missing=()
for id in T-A T-B T-C T-D T-E T-F T-G T-H T-I T-J; do
[[ -s "reports/${id}_results.xml" ]] || missing+=("$id")
done
if (( ${#missing[@]} )); then
echo "::error::Still missing T fragments: ${missing[*]}"
exit 1
fi
# (kept) Finalize staged report fragments (promote to reports/)
# (removed duplicate) Finalize staged report fragments
- name: Assert T coverage (after promotion)
if: always()
shell: bash
run: |
set -euo pipefail
missing=()
for id in T-A T-B T-C T-D T-E T-F T-G T-H T-I T-J; do
if [[ ! -s "reports/${id}_results.xml" ]]; then
# Accept staged fragment as present
[[ -s "reports/_staging/${id}_results.xml" ]] || missing+=("$id")
fi
done
if (( ${#missing[@]} )); then
echo "::error::Missing T fragments: ${missing[*]}"
exit 1
fi
- name: Canonicalize testcase names (NL/T prefixes)
if: always()
shell: bash
run: |
python3 - <<'PY'
from pathlib import Path
import xml.etree.ElementTree as ET, re, os
RULES = [
("NL-0", r"\b(NL-0|Baseline|State\s*Capture)\b"),
("NL-1", r"\b(NL-1|Core\s*Method)\b"),
("NL-2", r"\b(NL-2|Anchor|Build\s*marker)\b"),
("NL-3", r"\b(NL-3|End[-\s]*of[-\s]*Class\s*Content|Tail\s*test\s*[ABC])\b"),
("NL-4", r"\b(NL-4|Console|Unity\s*console)\b"),
("T-A", r"\b(T-?A|Temporary\s*Helper)\b"),
("T-B", r"\b(T-?B|Method\s*Body\s*Interior)\b"),
("T-C", r"\b(T-?C|Different\s*Method\s*Interior|ApplyBlend)\b"),
("T-D", r"\b(T-?D|End[-\s]*of[-\s]*Class\s*Helper|TestHelper)\b"),
("T-E", r"\b(T-?E|Method\s*Evolution|Counter|IncrementCounter)\b"),
("T-F", r"\b(T-?F|Atomic\s*Multi[-\s]*Edit)\b"),
("T-G", r"\b(T-?G|Path\s*Normalization)\b"),
("T-H", r"\b(T-?H|Validation\s*on\s*Modified)\b"),
("T-I", r"\b(T-?I|Failure\s*Surface)\b"),
("T-J", r"\b(T-?J|Idempotenc(y|e))\b"),
]
def canon_name(name: str) -> str:
n = name or ""
for tid, pat in RULES:
if re.search(pat, n, flags=re.I):
# If it already starts with the correct format, leave it alone
if re.match(rf'^\s*{re.escape(tid)}\s*[—–-]', n, flags=re.I):
return n.strip()
# If it has a different separator, extract title and reformat
title_match = re.search(rf'{re.escape(tid)}\s*[:.\-–—]\s*(.+)', n, flags=re.I)
if title_match:
title = title_match.group(1).strip()
return f"{tid} — {title}"
# Otherwise, just return the canonical ID
return tid
return n
def id_from_filename(p: Path):
n = p.name
m = re.match(r'NL(\d+)_results\.xml$', n, re.I)
if m:
return f"NL-{int(m.group(1))}"
m = re.match(r'T([A-J])_results\.xml$', n, re.I)
if m:
return f"T-{m.group(1).upper()}"
return None
frags = list(sorted(Path("reports").glob("*_results.xml")))
for frag in frags:
try:
tree = ET.parse(frag); root = tree.getroot()
except Exception:
continue
if root.tag != "testcase":
continue
file_id = id_from_filename(frag)
old = root.get("name") or ""
# Prefer filename-derived ID; if name doesn't start with it, override
if file_id:
# Respect file's ID (prevents T-D being renamed to NL-3 by loose patterns)
title = re.sub(r'^\s*(NL-\d+|T-[A-Z])\s*[—–:\-]\s*', '', old).strip()
new = f"{file_id} — {title}" if title else file_id
else:
new = canon_name(old)
if new != old and new:
root.set("name", new)
tree.write(frag, encoding="utf-8", xml_declaration=False)
print(f'canon: {frag.name}: "{old}" -> "{new}"')
# Note: Do not auto-relable fragments. We rely on per-test strict emission
# and the backfill step to surface missing tests explicitly.
PY
- name: Backfill missing NL/T tests (fail placeholders)
if: always()
shell: bash
run: |
python3 - <<'PY'
from pathlib import Path
import xml.etree.ElementTree as ET
import re
DESIRED = ["NL-0","NL-1","NL-2","NL-3","NL-4","T-A","T-B","T-C","T-D","T-E","T-F","T-G","T-H","T-I","T-J"]
seen = set()
def id_from_filename(p: Path):
n = p.name
m = re.match(r'NL(\d+)_results\.xml$', n, re.I)
if m:
return f"NL-{int(m.group(1))}"
m = re.match(r'T([A-J])_results\.xml$', n, re.I)
if m:
return f"T-{m.group(1).upper()}"
return None
for p in Path("reports").glob("*_results.xml"):
try:
r = ET.parse(p).getroot()
except Exception:
continue
# Count by filename id primarily; fall back to testcase name if needed
fid = id_from_filename(p)
if fid in DESIRED:
seen.add(fid)
continue
if r.tag == "testcase":
name = (r.get("name") or "").strip()
for d in DESIRED:
if name.startswith(d):
seen.add(d)
break
Path("reports").mkdir(parents=True, exist_ok=True)
for d in DESIRED:
if d in seen:
continue
frag = Path(f"reports/{d}_results.xml")
tc = ET.Element("testcase", {"classname":"UnityMCP.NL-T", "name": d})
fail = ET.SubElement(tc, "failure", {"message":"not produced"})
fail.text = "The agent did not emit a fragment for this test."
ET.ElementTree(tc).write(frag, encoding="utf-8", xml_declaration=False)
print(f"backfill: {d}")
PY
- name: "Debug: list testcase names"
if: always()
run: |
python3 - <<'PY'
from pathlib import Path
import xml.etree.ElementTree as ET
for p in sorted(Path('reports').glob('*_results.xml')):
try:
r = ET.parse(p).getroot()
if r.tag == 'testcase':
print(f"{p.name}: {(r.get('name') or '').strip()}")
except Exception:
pass
PY
# ---------- Merge testcase fragments into JUnit ---------- # ---------- Merge testcase fragments into JUnit ----------
- name: Normalize/assemble JUnit in-place (single file) - name: Normalize/assemble JUnit in-place (single file)
if: always() if: always()
@ -301,44 +646,96 @@ jobs:
from pathlib import Path from pathlib import Path
import xml.etree.ElementTree as ET import xml.etree.ElementTree as ET
import re, os import re, os
def localname(tag: str) -> str: return tag.rsplit('}', 1)[-1] if '}' in tag else tag
def localname(tag: str) -> str:
return tag.rsplit('}', 1)[-1] if '}' in tag else tag
src = Path(os.environ.get('JUNIT_OUT', 'reports/junit-nl-suite.xml')) src = Path(os.environ.get('JUNIT_OUT', 'reports/junit-nl-suite.xml'))
if not src.exists(): raise SystemExit(0) if not src.exists():
tree = ET.parse(src); root = tree.getroot() raise SystemExit(0)
tree = ET.parse(src)
root = tree.getroot()
suite = root.find('./*') if localname(root.tag) == 'testsuites' else root suite = root.find('./*') if localname(root.tag) == 'testsuites' else root
if suite is None: raise SystemExit(0) if suite is None:
raise SystemExit(0)
def id_from_filename(p: Path):
n = p.name
m = re.match(r'NL(\d+)_results\.xml$', n, re.I)
if m:
return f"NL-{int(m.group(1))}"
m = re.match(r'T([A-J])_results\.xml$', n, re.I)
if m:
return f"T-{m.group(1).upper()}"
return None
def id_from_system_out(tc):
so = tc.find('system-out')
if so is not None and so.text:
m = re.search(r'\b(NL-\d+|T-[A-Z])\b', so.text)
if m:
return m.group(1)
return None
fragments = sorted(Path('reports').glob('*_results.xml')) fragments = sorted(Path('reports').glob('*_results.xml'))
added = 0 added = 0
renamed = 0
for frag in fragments: for frag in fragments:
tcs = []
try: try:
froot = ET.parse(frag).getroot() froot = ET.parse(frag).getroot()
if localname(froot.tag) == 'testcase': if localname(froot.tag) == 'testcase':
suite.append(froot); added += 1 tcs = [froot]
else: else:
for tc in froot.findall('.//testcase'): tcs = list(froot.findall('.//testcase'))
suite.append(tc); added += 1
except Exception: except Exception:
txt = Path(frag).read_text(encoding='utf-8', errors='replace') txt = Path(frag).read_text(encoding='utf-8', errors='replace')
for m in re.findall(r'<testcase[\\s\\S]*?</testcase>', txt, flags=re.DOTALL): # Extract all testcase nodes from raw text
try: suite.append(ET.fromstring(m)); added += 1 nodes = re.findall(r'<testcase[\s\S]*?</testcase>', txt, flags=re.DOTALL)
except Exception: pass for m in nodes:
try:
tcs.append(ET.fromstring(m))
except Exception:
pass
# Guard: keep only the first testcase from each fragment
if len(tcs) > 1:
tcs = tcs[:1]
test_id = id_from_filename(frag)
for tc in tcs:
current_name = tc.get('name') or ''
tid = test_id or id_from_system_out(tc)
# Enforce filename-derived ID as prefix; repair names if needed
if tid and not re.match(r'^\s*(NL-\d+|T-[A-Z])\b', current_name):
title = current_name.strip()
new_name = f'{tid} — {title}' if title else tid
tc.set('name', new_name)
elif tid and not re.match(rf'^\s*{re.escape(tid)}\b', current_name):
# Replace any wrong leading ID with the correct one
title = re.sub(r'^\s*(NL-\d+|T-[A-Z])\s*[—–:\-]\s*', '', current_name).strip()
new_name = f'{tid} — {title}' if title else tid
tc.set('name', new_name)
renamed += 1
suite.append(tc)
added += 1
if added: if added:
# Drop bootstrap placeholder and recompute counts # Drop bootstrap placeholder and recompute counts
removed_bootstrap = 0
for tc in list(suite.findall('.//testcase')): for tc in list(suite.findall('.//testcase')):
name = (tc.get('name') or '') if (tc.get('name') or '') == 'NL-Suite.Bootstrap':
if name == 'NL-Suite.Bootstrap':
suite.remove(tc) suite.remove(tc)
removed_bootstrap += 1
testcases = suite.findall('.//testcase') testcases = suite.findall('.//testcase')
tests_cnt = len(testcases)
failures_cnt = sum(1 for tc in testcases if (tc.find('failure') is not None or tc.find('error') is not None)) failures_cnt = sum(1 for tc in testcases if (tc.find('failure') is not None or tc.find('error') is not None))
suite.set('tests', str(tests_cnt)) suite.set('tests', str(len(testcases)))
suite.set('failures', str(failures_cnt)) suite.set('failures', str(failures_cnt))
suite.set('errors', str(0)) suite.set('errors', '0')
suite.set('skipped', str(0)) suite.set('skipped', '0')
tree.write(src, encoding='utf-8', xml_declaration=True) tree.write(src, encoding='utf-8', xml_declaration=True)
print(f"Added {added} testcase fragments; removed bootstrap={removed_bootstrap}; tests={tests_cnt}; failures={failures_cnt}") print(f"Appended {added} testcase(s); renamed {renamed} to canonical NL/T names.")
PY PY
# ---------- Markdown summary from JUnit ---------- # ---------- Markdown summary from JUnit ----------
@ -349,14 +746,13 @@ jobs:
python3 - <<'PY' python3 - <<'PY'
import xml.etree.ElementTree as ET import xml.etree.ElementTree as ET
from pathlib import Path from pathlib import Path
import os, html import os, html, re
def localname(tag: str) -> str: def localname(tag: str) -> str:
return tag.rsplit('}', 1)[-1] if '}' in tag else tag return tag.rsplit('}', 1)[-1] if '}' in tag else tag
src = Path(os.environ.get('JUNIT_OUT', 'reports/junit-nl-suite.xml')) src = Path(os.environ.get('JUNIT_OUT', 'reports/junit-nl-suite.xml'))
md_out = Path(os.environ.get('MD_OUT', 'reports/junit-nl-suite.md')) md_out = Path(os.environ.get('MD_OUT', 'reports/junit-nl-suite.md'))
# Ensure destination directory exists even if earlier prep steps were skipped
md_out.parent.mkdir(parents=True, exist_ok=True) md_out.parent.mkdir(parents=True, exist_ok=True)
if not src.exists(): if not src.exists():
@ -368,19 +764,33 @@ jobs:
suite = root.find('./*') if localname(root.tag) == 'testsuites' else root suite = root.find('./*') if localname(root.tag) == 'testsuites' else root
cases = [] if suite is None else list(suite.findall('.//testcase')) cases = [] if suite is None else list(suite.findall('.//testcase'))
def id_from_case(tc):
n = (tc.get('name') or '')
m = re.match(r'\s*(NL-\d+|T-[A-Z])\b', n)
if m:
return m.group(1)
so = tc.find('system-out')
if so is not None and so.text:
m = re.search(r'\b(NL-\d+|T-[A-Z])\b', so.text)
if m:
return m.group(1)
return None
id_status = {}
name_map = {}
for tc in cases:
tid = id_from_case(tc)
ok = (tc.find('failure') is None and tc.find('error') is None)
if tid and tid not in id_status:
id_status[tid] = ok
name_map[tid] = (tc.get('name') or tid)
desired = ['NL-0','NL-1','NL-2','NL-3','NL-4','T-A','T-B','T-C','T-D','T-E','T-F','T-G','T-H','T-I','T-J']
total = len(cases) total = len(cases)
failures = sum(1 for tc in cases if (tc.find('failure') is not None or tc.find('error') is not None)) failures = sum(1 for tc in cases if (tc.find('failure') is not None or tc.find('error') is not None))
passed = total - failures passed = total - failures
desired = ['NL-0','NL-1','NL-2','NL-3','NL-4','T-A','T-B','T-C','T-D','T-E','T-F','T-G','T-H','T-I','T-J']
name_to_case = {(tc.get('name') or ''): tc for tc in cases}
def status_for(prefix: str):
for name, tc in name_to_case.items():
if name.startswith(prefix):
return not ((tc.find('failure') is not None) or (tc.find('error') is not None))
return None
lines = [] lines = []
lines += [ lines += [
'# Unity NL/T Editing Suite Test Results', '# Unity NL/T Editing Suite Test Results',
@ -390,52 +800,59 @@ jobs:
'## Test Checklist' '## Test Checklist'
] ]
for p in desired: for p in desired:
st = status_for(p) st = id_status.get(p, None)
lines.append(f"- [x] {p}" if st is True else (f"- [ ] {p} (fail)" if st is False else f"- [ ] {p} (not run)")) lines.append(f"- [x] {p}" if st is True else (f"- [ ] {p} (fail)" if st is False else f"- [ ] {p} (not run)"))
lines.append('') lines.append('')
# Rich per-test system-out details
lines.append('## Test Details') lines.append('## Test Details')
def order_key(n: str): def order_key(n: str):
if n.startswith('NL-'):
try: try:
if n.startswith('NL-') and n[3].isdigit(): return (0, int(n.split('-')[1]))
return (0, int(n.split('.')[0].split('-')[1])) except:
except Exception: return (0, 999)
pass if n.startswith('T-') and len(n) > 2:
if n.startswith('T-') and len(n) > 2 and n[2].isalpha():
return (1, ord(n[2])) return (1, ord(n[2]))
return (2, n) return (2, n)
MAX_CHARS = 2000 MAX_CHARS = 2000
for name in sorted(name_to_case.keys(), key=order_key): seen = set()
tc = name_to_case[name] for tid in sorted(id_status.keys(), key=order_key):
status_badge = "PASS" if (tc.find('failure') is None and tc.find('error') is None) else "FAIL" seen.add(tid)
lines.append(f"### {name} — {status_badge}") tc = next((c for c in cases if (id_from_case(c) == tid)), None)
if not tc:
continue
title = name_map.get(tid, tid)
status_badge = "PASS" if id_status[tid] else "FAIL"
lines.append(f"### {title} — {status_badge}")
so = tc.find('system-out') so = tc.find('system-out')
text = '' if so is None or so.text is None else so.text.replace('\r\n','\n') text = '' if so is None or so.text is None else html.unescape(so.text.replace('\r\n','\n'))
# Unescape XML entities so code reads naturally (e.g., => instead of =&gt;)
if text:
text = html.unescape(text)
if text.strip(): if text.strip():
t = text.strip() t = text.strip()
if len(t) > MAX_CHARS: if len(t) > MAX_CHARS:
t = t[:MAX_CHARS] + "\n…(truncated)" t = t[:MAX_CHARS] + "\n…(truncated)"
# Use a safer fence if content contains triple backticks fence = '```' if '```' not in t else '````'
fence = '```' lines += [fence, t, fence]
if '```' in t:
fence = '````'
lines.append(fence)
lines.append(t)
lines.append(fence)
else: else:
lines.append('(no system-out)') lines.append('(no system-out)')
node = tc.find('failure') or tc.find('error') node = tc.find('failure') or tc.find('error')
if node is not None: if node is not None:
msg = (node.get('message') or '').strip() msg = (node.get('message') or '').strip()
body = (node.text or '').strip() body = (node.text or '').strip()
if msg: lines.append(f"- Message: {msg}") if msg:
if body: lines.append(f"- Detail: {body.splitlines()[0][:500]}") lines.append(f"- Message: {msg}")
if body:
lines.append(f"- Detail: {body.splitlines()[0][:500]}")
lines.append('')
for tc in cases:
if id_from_case(tc) in seen:
continue
title = tc.get('name') or '(unnamed)'
status_badge = "PASS" if (tc.find('failure') is None and tc.find('error') is None) else "FAIL"
lines.append(f"### {title} — {status_badge}")
lines.append('(unmapped test id)')
lines.append('') lines.append('')
md_out.write_text('\n'.join(lines), encoding='utf-8') md_out.write_text('\n'.join(lines), encoding='utf-8')
@ -478,7 +895,7 @@ jobs:
p.write_text(s, encoding='utf-8', newline='\n') p.write_text(s, encoding='utf-8', newline='\n')
PY PY
- name: NL/T details → Job Summary - name: NL/T details -> Job Summary
if: always() if: always()
run: | run: |
echo "## Unity NL/T Editing Suite — Summary" >> $GITHUB_STEP_SUMMARY echo "## Unity NL/T Editing Suite — Summary" >> $GITHUB_STEP_SUMMARY
@ -538,6 +955,15 @@ jobs:
- name: Stop Unity - name: Stop Unity
if: always() if: always()
run: | run: |
docker logs --tail 400 unity-mcp | sed -E 's/((serial|license|password|token)[^[:space:]]*)/[REDACTED]/ig' || true docker logs --tail 400 unity-mcp | sed -E 's/((email|serial|license|password|token)[^[:space:]]*)/[REDACTED]/ig' || true
docker rm -f unity-mcp || true docker rm -f unity-mcp || true
- name: Return Pro license (if used)
if: always() && steps.lic.outputs.use_ebl == 'true' && steps.lic.outputs.has_serial == 'true'
uses: game-ci/unity-return-license@v2
continue-on-error: true
env:
UNITY_EMAIL: ${{ secrets.UNITY_EMAIL }}
UNITY_PASSWORD: ${{ secrets.UNITY_PASSWORD }}
UNITY_SERIAL: ${{ secrets.UNITY_SERIAL }}

113
README-DEV.md
View File

@ -16,6 +16,22 @@ Quick deployment and testing tools for MCP for Unity core changes.
--- ---
## Switching MCP package sources quickly
Run this from the unity-mcp repo, not your game's roote directory. Use `mcp_source.py` to quickly switch between different MCP for Unity package sources:
**Usage:**
```bash
python mcp_source.py [--manifest /path/to/manifest.json] [--repo /path/to/unity-mcp] [--choice 1|2|3]
```
**Options:**
- **1** Upstream main (CoplayDev/unity-mcp)
- **2** Remote current branch (origin + branch)
- **3** Local workspace (file: UnityMcpBridge)
After switching, open Package Manager and Refresh to re-resolve packages.
## Development Deployment Scripts ## Development Deployment Scripts
These deployment scripts help you quickly test changes to MCP for Unity core code. These deployment scripts help you quickly test changes to MCP for Unity core code.
@ -46,6 +62,18 @@ Restores original files from backup.
2. Allows you to select which backup to restore 2. Allows you to select which backup to restore
3. Restores both Unity Bridge and Python Server files 3. Restores both Unity Bridge and Python Server files
### `prune_tool_results.py`
Compacts large `tool_result` blobs in conversation JSON into concise one-line summaries.
**Usage:**
```bash
python3 prune_tool_results.py < reports/claude-execution-output.json > reports/claude-execution-output.pruned.json
```
The script reads a conversation from `stdin` and writes the pruned version to `stdout`, making logs much easier to inspect or archive.
These defaults dramatically cut token usage without affecting essential information.
## Finding Unity Package Cache Path ## Finding Unity Package Cache Path
Unity stores Git packages under a version-or-hash folder. Expect something like: Unity stores Git packages under a version-or-hash folder. Expect something like:
@ -66,12 +94,67 @@ To find it reliably:
Note: In recent builds, the Python server sources are also bundled inside the package under `UnityMcpServer~/src`. This is handy for local testing or pointing MCP clients directly at the packaged server. Note: In recent builds, the Python server sources are also bundled inside the package under `UnityMcpServer~/src`. This is handy for local testing or pointing MCP clients directly at the packaged server.
## MCP Bridge Stress Test
An on-demand stress utility exercises the MCP bridge with multiple concurrent clients while triggering real script reloads via immediate script edits (no menu calls required).
### Script
- `tools/stress_mcp.py`
### What it does
- Starts N TCP clients against the Unity MCP bridge (default port auto-discovered from `~/.unity-mcp/unity-mcp-status-*.json`).
- Sends lightweight framed `ping` keepalives to maintain concurrency.
- In parallel, appends a unique marker comment to a target C# file using `manage_script.apply_text_edits` with:
- `options.refresh = "immediate"` to force an import/compile immediately (triggers domain reload), and
- `precondition_sha256` computed from the current file contents to avoid drift.
- Uses EOF insertion to avoid header/`using`-guard edits.
### Usage (local)
```bash
# Recommended: use the included large script in the test project
python3 tools/stress_mcp.py \
--duration 60 \
--clients 8 \
--unity-file "TestProjects/UnityMCPTests/Assets/Scripts/LongUnityScriptClaudeTest.cs"
```
Flags:
- `--project` Unity project path (auto-detected to the included test project by default)
- `--unity-file` C# file to edit (defaults to the long test script)
- `--clients` number of concurrent clients (default 10)
- `--duration` seconds to run (default 60)
### Expected outcome
- No Unity Editor crashes during reload churn
- Immediate reloads after each applied edit (no `Assets/Refresh` menu calls)
- Some transient disconnects or a few failed calls may occur during domain reload; the tool retries and continues
- JSON summary printed at the end, e.g.:
- `{"port": 6400, "stats": {"pings": 28566, "applies": 69, "disconnects": 0, "errors": 0}}`
### Notes and troubleshooting
- Immediate vs debounced:
- The tool sets `options.refresh = "immediate"` so changes compile instantly. If you only need churn (not per-edit confirmation), switch to debounced to reduce mid-reload failures.
- Precondition required:
- `apply_text_edits` requires `precondition_sha256` on larger files. The tool reads the file first to compute the SHA.
- Edit location:
- To avoid header guards or complex ranges, the tool appends a one-line marker at EOF each cycle.
- Read API:
- The bridge currently supports `manage_script.read` for file reads. You may see a deprecation warning; it's harmless for this internal tool.
- Transient failures:
- Occasional `apply_errors` often indicate the connection reloaded mid-reply. Edits still typically apply; the loop continues on the next iteration.
### CI guidance
- Keep this out of default PR CI due to Unity/editor requirements and runtime variability.
- Optionally run it as a manual workflow or nightly job on a Unity-capable runner.
## CI Test Workflow (GitHub Actions) ## CI Test Workflow (GitHub Actions)
We provide a CI job to run a Natural Language Editing mini-suite against the Unity test project. It spins up a headless Unity container and connects via the MCP bridge. We provide a CI job to run a Natural Language Editing suite against the Unity test project. It spins up a headless Unity container and connects via the MCP bridge. To run from your fork, you need the following GitHub "secrets": an `ANTHROPIC_API_KEY` and Unity credentials (usually `UNITY_EMAIL` + `UNITY_PASSWORD` or `UNITY_LICENSE` / `UNITY_SERIAL`.) These are redacted in logs so never visible.
- Trigger: Workflow dispatch (`Claude NL suite (Unity live)`). ***To run it***
- Trigger: In GitHun "Actions" for the repo, trigger `workflow dispatch` (`Claude NL/T Full Suite (Unity live)`).
- Image: `UNITY_IMAGE` (UnityCI) pulled by tag; the job resolves a digest at runtime. Logs are sanitized. - Image: `UNITY_IMAGE` (UnityCI) pulled by tag; the job resolves a digest at runtime. Logs are sanitized.
- Execution: single pass with immediate per‑test fragment emissions (strict single `<testcase>` per file). A placeholder guard fails fast if any fragment is a bare ID. Staging (`reports/_staging`) is promoted to `reports/` to reduce partial writes.
- Reports: JUnit at `reports/junit-nl-suite.xml`, Markdown at `reports/junit-nl-suite.md`. - Reports: JUnit at `reports/junit-nl-suite.xml`, Markdown at `reports/junit-nl-suite.md`.
- Publishing: JUnit is normalized to `reports/junit-for-actions.xml` and published; artifacts upload all files under `reports/`. - Publishing: JUnit is normalized to `reports/junit-for-actions.xml` and published; artifacts upload all files under `reports/`.
@ -80,10 +163,9 @@ We provide a CI job to run a Natural Language Editing mini-suite against the Uni
- `TestProjects/UnityMCPTests/Assets/Scripts/LongUnityScriptClaudeTest.cs` - `TestProjects/UnityMCPTests/Assets/Scripts/LongUnityScriptClaudeTest.cs`
Use this file locally and in CI to validate multi-edit batches, anchor inserts, and windowed reads on a sizable script. Use this file locally and in CI to validate multi-edit batches, anchor inserts, and windowed reads on a sizable script.
### Add a new NL test ### Adjust tests / prompts
- Edit `.claude/prompts/nl-unity-claude-tests-mini.md` (or `nl-unity-suite-full.md` for the larger suite). - Edit `.claude/prompts/nl-unity-suite-t.md` to modify the NL/T steps. Follow the conventions: emit one XML fragment per test under `reports/<TESTID>_results.xml`, each containing exactly one `<testcase>` with a `name` that begins with the test ID. No prologue/epilogue or code fences.
- Follow the conventions: single `<testsuite>` root, one `<testcase>` per sub-test, end system-out with `VERDICT: PASS|FAIL`. - Keep edits minimal and reversible; include concise evidence.
- Keep edits minimal and reversible; include evidence windows and compact diffs.
### Run the suite ### Run the suite
1) Push your branch, then manually run the workflow from the Actions tab. 1) Push your branch, then manually run the workflow from the Actions tab.
@ -95,7 +177,6 @@ We provide a CI job to run a Natural Language Editing mini-suite against the Uni
- Check: “JUnit Test Report” on the PR/commit. - Check: “JUnit Test Report” on the PR/commit.
- Artifacts: `claude-nl-suite-artifacts` includes XML and MD. - Artifacts: `claude-nl-suite-artifacts` includes XML and MD.
### MCP Connection Debugging ### MCP Connection Debugging
- *Enable debug logs* in the Unity MCP window (inside the Editor) to view connection status, auto-setup results, and MCP client paths. It shows: - *Enable debug logs* in the Unity MCP window (inside the Editor) to view connection status, auto-setup results, and MCP client paths. It shows:
- bridge startup/port, client connections, strict framing negotiation, and parsed frames - bridge startup/port, client connections, strict framing negotiation, and parsed frames
@ -109,24 +190,6 @@ We provide a CI job to run a Natural Language Editing mini-suite against the Uni
4. **Iterate** - repeat steps 1-3 as needed 4. **Iterate** - repeat steps 1-3 as needed
5. **Restore** original files when done using `restore-dev.bat` 5. **Restore** original files when done using `restore-dev.bat`
## Switching MCP package sources quickly
Use `mcp_source.py` to quickly switch between different MCP for Unity package sources:
**Usage:**
```bash
python mcp_source.py [--manifest /path/to/manifest.json] [--repo /path/to/unity-mcp] [--choice 1|2|3]
```
**Options:**
- **1** Upstream main (CoplayDev/unity-mcp)
- **2** Remote current branch (origin + branch)
- **3** Local workspace (file: UnityMcpBridge)
After switching, open Package Manager and Refresh to re-resolve packages.
## Troubleshooting ## Troubleshooting
### "Path not found" errors running the .bat file ### "Path not found" errors running the .bat file

154
README.md
View File

@ -1,4 +1,4 @@
# MCP for Unity ✨ <img width="676" height="380" alt="MCP for Unity" src="https://github.com/user-attachments/assets/b712e41d-273c-48b2-9041-82bd17ace267" />
#### Proudly sponsored and maintained by [Coplay](https://www.coplay.dev/?ref=unity-mcp) -- the best AI assistant for Unity. [Read the backstory here.](https://www.coplay.dev/blog/coplay-and-open-source-unity-mcp-join-forces) #### Proudly sponsored and maintained by [Coplay](https://www.coplay.dev/?ref=unity-mcp) -- the best AI assistant for Unity. [Read the backstory here.](https://www.coplay.dev/blog/coplay-and-open-source-unity-mcp-join-forces)
@ -15,9 +15,9 @@
MCP for Unity acts as a bridge, allowing AI assistants (like Claude, Cursor) to interact directly with your Unity Editor via a local **MCP (Model Context Protocol) Client**. Give your LLM tools to manage assets, control scenes, edit scripts, and automate tasks within Unity. MCP for Unity acts as a bridge, allowing AI assistants (like Claude, Cursor) to interact directly with your Unity Editor via a local **MCP (Model Context Protocol) Client**. Give your LLM tools to manage assets, control scenes, edit scripts, and automate tasks within Unity.
## 💬 Join Our Community ---
### [Discord](https://discord.gg/y4p8KfzrN4) ### 💬 Join Our [Discord](https://discord.gg/y4p8KfzrN4)
**Get help, share ideas, and collaborate with other MCP for Unity developers!** **Get help, share ideas, and collaborate with other MCP for Unity developers!**
@ -37,7 +37,7 @@ MCP for Unity acts as a bridge, allowing AI assistants (like Claude, Cursor) to
* `read_console`: Gets messages from or clears the console. * `read_console`: Gets messages from or clears the console.
* `manage_script`: Manages C# scripts (create, read, update, delete). * `manage_script`: Manages C# scripts (create, read, update, delete).
* `manage_editor`: Controls and queries the editor\'s state and settings. * `manage_editor`: Controls and queries the editor's state and settings.
* `manage_scene`: Manages scenes (load, save, create, get hierarchy, etc.). * `manage_scene`: Manages scenes (load, save, create, get hierarchy, etc.).
* `manage_asset`: Performs asset operations (import, create, modify, delete, etc.). * `manage_asset`: Performs asset operations (import, create, modify, delete, etc.).
* `manage_shader`: Performs shader CRUD operations (create, read, modify, delete). * `manage_shader`: Performs shader CRUD operations (create, read, modify, delete).
@ -50,37 +50,36 @@ MCP for Unity acts as a bridge, allowing AI assistants (like Claude, Cursor) to
--- ---
## How It Works 🤔 ## How It Works
MCP for Unity connects your tools using two components: MCP for Unity connects your tools using two components:
1. **MCP for Unity Bridge:** A Unity package running inside the Editor. (Installed via Package Manager). 1. **MCP for Unity Bridge:** A Unity package running inside the Editor. (Installed via Package Manager).
2. **MCP for Unity Server:** A Python server that runs locally, communicating between the Unity Bridge and your MCP Client. (Installed automatically by the package on first run or via Auto-Setup; manual setup is available as a fallback). 2. **MCP for Unity Server:** A Python server that runs locally, communicating between the Unity Bridge and your MCP Client. (Installed automatically by the package on first run or via Auto-Setup; manual setup is available as a fallback).
**Flow:** `[Your LLM via MCP Client] <-> [MCP for Unity Server (Python)] <-> [MCP for Unity Bridge (Unity Editor)]` <img width="562" height="121" alt="image" src="https://github.com/user-attachments/assets/9abf9c66-70d1-4b82-9587-658e0d45dc3e" />
--- ---
## Installation ⚙️ ## Installation ⚙️
> **Note:** The setup is constantly improving as we update the package. Check back if you randomly start to run into issues.
### Prerequisites ### Prerequisites
* **Python:** Version 3.12 or newer. [Download Python](https://www.python.org/downloads/) * **Python:** Version 3.12 or newer. [Download Python](https://www.python.org/downloads/)
* **Unity Hub & Editor:** Version 2021.3 LTS or newer. [Download Unity](https://unity.com/download) * **Unity Hub & Editor:** Version 2021.3 LTS or newer. [Download Unity](https://unity.com/download)
* **uv (Python package manager):** * **uv (Python toolchain manager):**
```bash ```bash
pip install uv # macOS / Linux
# Or see: https://docs.astral.sh/uv/getting-started/installation/ curl -LsSf https://astral.sh/uv/install.sh | sh
# Windows (PowerShell)
winget install Astral.Sh.Uv
# Docs: https://docs.astral.sh/uv/getting-started/installation/
``` ```
* **An MCP Client:**
* [Claude Desktop](https://claude.ai/download) * **An MCP Client:** : [Claude Desktop](https://claude.ai/download) | [Claude Code](https://github.com/anthropics/claude-code) | [Cursor](https://www.cursor.com/en/downloads) | [Visual Studio Code Copilot](https://code.visualstudio.com/docs/copilot/overview) | [Windsurf](https://windsurf.com) | Others work with manual config
* [Claude Code](https://github.com/anthropics/claude-code)
* [Cursor](https://www.cursor.com/en/downloads)
* [Visual Studio Code Copilot](https://code.visualstudio.com/docs/copilot/overview)
* [Windsurf](https://windsurf.com)
* *(Others may work with manual config)*
* <details> <summary><strong>[Optional] Roslyn for Advanced Script Validation</strong></summary> * <details> <summary><strong>[Optional] Roslyn for Advanced Script Validation</strong></summary>
For **Strict** validation level that catches undefined namespaces, types, and methods: For **Strict** validation level that catches undefined namespaces, types, and methods:
@ -102,7 +101,8 @@ MCP for Unity connects your tools using two components:
**Note:** Without Roslyn, script validation falls back to basic structural checks. Roslyn enables full C# compiler diagnostics with precise error reporting.</details> **Note:** Without Roslyn, script validation falls back to basic structural checks. Roslyn enables full C# compiler diagnostics with precise error reporting.</details>
### 🌟Step 1: Install the Unity Package🌟 ---
### 🌟 Step 1: Install the Unity Package
#### To install via Git URL #### To install via Git URL
@ -118,14 +118,13 @@ MCP for Unity connects your tools using two components:
#### To install via OpenUPM #### To install via OpenUPM
1. Instal the [OpenUPM CLI](https://openupm.com/docs/getting-started-cli.html) 1. Install the [OpenUPM CLI](https://openupm.com/docs/getting-started-cli.html)
2. Open a terminal (PowerShell, Terminal, etc.) and navigate to your Unity project directory 2. Open a terminal (PowerShell, Terminal, etc.) and navigate to your Unity project directory
3. Run `openupm add com.coplaydev.unity-mcp` 3. Run `openupm add com.coplaydev.unity-mcp`
**Note:** If you installed the MCP Server before Coplay's maintenance, you will need to uninstall the old package before re-installing the new one. **Note:** If you installed the MCP Server before Coplay's maintenance, you will need to uninstall the old package before re-installing the new one.
### Step 2: Configure Your MCP Client ### 🛠️ Step 2: Configure Your MCP Client
Connect your MCP Client (Claude, Cursor, etc.) to the Python server set up in Step 1 (auto) or via Manual Configuration (below). Connect your MCP Client (Claude, Cursor, etc.) to the Python server set up in Step 1 (auto) or via Manual Configuration (below).
<img width="648" height="599" alt="MCPForUnity-Readme-Image" src="https://github.com/user-attachments/assets/b4a725da-5c43-4bd6-80d6-ee2e3cca9596" /> <img width="648" height="599" alt="MCPForUnity-Readme-Image" src="https://github.com/user-attachments/assets/b4a725da-5c43-4bd6-80d6-ee2e3cca9596" />
@ -134,7 +133,7 @@ Connect your MCP Client (Claude, Cursor, etc.) to the Python server set up in St
1. In Unity, go to `Window > MCP for Unity`. 1. In Unity, go to `Window > MCP for Unity`.
2. Click `Auto-Setup`. 2. Click `Auto-Setup`.
3. Look for a green status indicator 🟢 and "Connected ✓". *(This attempts to modify the MCP Client\'s config file automatically).* 3. Look for a green status indicator 🟢 and "Connected ✓". *(This attempts to modify the MCP Client's config file automatically).*
<details><summary><strong>Client-specific troubleshooting</strong></summary> <details><summary><strong>Client-specific troubleshooting</strong></summary>
@ -147,7 +146,7 @@ Connect your MCP Client (Claude, Cursor, etc.) to the Python server set up in St
If Auto-Setup fails or you use a different client: If Auto-Setup fails or you use a different client:
1. **Find your MCP Client\'s configuration file.** (Check client documentation). 1. **Find your MCP Client's configuration file.** (Check client documentation).
* *Claude Example (macOS):* `~/Library/Application Support/Claude/claude_desktop_config.json` * *Claude Example (macOS):* `~/Library/Application Support/Claude/claude_desktop_config.json`
* *Claude Example (Windows):* `%APPDATA%\Claude\claude_desktop_config.json` * *Claude Example (Windows):* `%APPDATA%\Claude\claude_desktop_config.json`
2. **Edit the file** to add/update the `mcpServers` section, using the *exact* paths from Step 1. 2. **Edit the file** to add/update the `mcpServers` section, using the *exact* paths from Step 1.
@ -155,6 +154,23 @@ If Auto-Setup fails or you use a different client:
<details> <details>
<summary><strong>Click for Client-Specific JSON Configuration Snippets...</strong></summary> <summary><strong>Click for Client-Specific JSON Configuration Snippets...</strong></summary>
---
**Claude Code**
If you're using Claude Code, you can register the MCP server using the below commands:
🚨**make sure to run these from your Unity project's home directory**🚨
**macOS:**
```bash
claude mcp add UnityMCP -- uv --directory /Users/USERNAME/Library/AppSupport/UnityMCP/UnityMcpServer/src run server.py
```
**Windows:**
```bash
claude mcp add UnityMCP -- "C:/Users/USERNAME/AppData/Local/Microsoft/WinGet/Links/uv.exe" --directory "C:/Users/USERNAME/AppData/Local/UnityMCP/UnityMcpServer/src" run server.py
```
**VSCode (all OS)** **VSCode (all OS)**
```json ```json
@ -181,7 +197,7 @@ On Windows, set `command` to the absolute shim, e.g. `C:\\Users\\YOU\\AppData\\L
"args": [ "args": [
"run", "run",
"--directory", "--directory",
"C:\\Users\\YOUR_USERNAME\\AppData\\Local\\Programs\\UnityMCP\\UnityMcpServer\\src", "C:\\Users\\YOUR_USERNAME\\AppData\\Local\\UnityMCP\\UnityMcpServer\\src",
"server.py" "server.py"
] ]
} }
@ -202,7 +218,7 @@ On Windows, set `command` to the absolute shim, e.g. `C:\\Users\\YOU\\AppData\\L
"args": [ "args": [
"run", "run",
"--directory", "--directory",
"/usr/local/bin/UnityMCP/UnityMcpServer/src", "/Users/YOUR_USERNAME/Library/AppSupport/UnityMCP/UnityMcpServer/src",
"server.py" "server.py"
] ]
} }
@ -211,7 +227,7 @@ On Windows, set `command` to the absolute shim, e.g. `C:\\Users\\YOU\\AppData\\L
} }
``` ```
(Replace YOUR_USERNAME if using ~/bin) (Replace YOUR_USERNAME. Note: AppSupport is a symlink to "Application Support" to avoid quoting issues)
**Linux:** **Linux:**
@ -223,7 +239,7 @@ On Windows, set `command` to the absolute shim, e.g. `C:\\Users\\YOU\\AppData\\L
"args": [ "args": [
"run", "run",
"--directory", "--directory",
"/home/YOUR_USERNAME/bin/UnityMCP/UnityMcpServer/src", "/home/YOUR_USERNAME/.local/share/UnityMCP/UnityMcpServer/src",
"server.py" "server.py"
] ]
} }
@ -234,21 +250,7 @@ On Windows, set `command` to the absolute shim, e.g. `C:\\Users\\YOU\\AppData\\L
(Replace YOUR_USERNAME) (Replace YOUR_USERNAME)
**For Claude Code**
If you\'re using Claude Code, you can register the MCP server using these commands:
**macOS:**
```bash
claude mcp add UnityMCP -- uv --directory /[PATH_TO]/UnityMCP/UnityMcpServer/src run server.py
```
**Windows:**
```bash
claude mcp add UnityMCP -- "C:/Users/USERNAME/AppData/Roaming/Python/Python313/Scripts/uv.exe" --directory "C:/Users/USERNAME/AppData/Local/Programs/UnityMCP/UnityMcpServer/src" run server.py
```
</details> </details>
--- ---
@ -261,58 +263,19 @@ claude mcp add UnityMCP -- "C:/Users/USERNAME/AppData/Roaming/Python/Python313/S
3. **Interact!** Unity tools should now be available in your MCP Client. 3. **Interact!** Unity tools should now be available in your MCP Client.
Example Prompt: `Create a 3D player controller`, `Create a yellow and bridge sun`, `Create a cool shader and apply it on a cube`. Example Prompt: `Create a 3D player controller`, `Create a tic-tac-toe game in 3D`, `Create a cool shader and apply to a cube`.
--- ---
## Future Dev Plans (Besides PR) 📝 ## Development & Contributing 🛠️
### 🔴 High Priority ### For Developers
- [ ] **Asset Generation Improvements** - Enhanced server request handling and asset pipeline optimization If you're contributing to MCP for Unity or want to test core changes, we have development tools to streamline your workflow:
- [ ] **Code Generation Enhancements** - Improved generated code quality and error handling
- [ ] **Robust Error Handling** - Comprehensive error messages, recovery mechanisms, and graceful degradation
- [ ] **Remote Connection Support** - Enable seamless remote connection between Unity host and MCP server
- [ ] **Documentation Expansion** - Complete tutorials for custom tool creation and API reference
### 🟡 Medium Priority
- [ ] **Custom Tool Creation GUI** - Visual interface for users to create and configure their own MCP tools
- [ ] **Advanced Logging System** - Logging with filtering, export, and debugging capabilities
### 🟢 Low Priority
- [ ] **Mobile Platform Support** - Extended toolset for mobile development workflows and platform-specific features
- [ ] **Easier Tool Setup**
- [ ] **Plugin Marketplace** - Community-driven tool sharing and distribution platform
<details open>
<summary><strong>✅ Completed Features<strong></summary>
- [x] **Shader Generation** - Generate shaders using CGProgram template
- [x] **Advanced Script Validation** - Multi-level validation with semantic analysis, namespace/type checking, and Unity best practices (Will need Roslyn Installed, see [Prerequisite](#prerequisites)).
</details>
### 🔬 Research & Exploration
- [ ] **AI-Powered Asset Generation** - Integration with AI tools for automatic 3D models, textures, and animations
- [ ] **Real-time Collaboration** - Live editing sessions between multiple developers *(Currently in progress)*
- [ ] **Analytics Dashboard** - Usage analytics, project insights, and performance metrics
- [ ] **Voice Commands** - Voice-controlled Unity operations for accessibility
- [ ] **AR/VR Tool Integration** - Extended support for immersive development workflows
---
## For Developers 🛠️
### Development Tools
If you\'re contributing to MCP for Unity or want to test core changes, we have development tools to streamline your workflow:
- **Development Deployment Scripts**: Quickly deploy and test your changes to MCP for Unity Bridge and Python Server - **Development Deployment Scripts**: Quickly deploy and test your changes to MCP for Unity Bridge and Python Server
- **Automatic Backup System**: Safe testing with easy rollback capabilities - **Automatic Backup System**: Safe testing with easy rollback capabilities
- **Hot Reload Workflow**: Fast iteration cycle for core development - **Hot Reload Workflow**: Fast iteration cycle for core development
- **More coming!**
📖 **See [README-DEV.md](README-DEV.md)** for complete development setup and workflow documentation. 📖 **See [README-DEV.md](README-DEV.md)** for complete development setup and workflow documentation.
@ -321,15 +284,10 @@ If you\'re contributing to MCP for Unity or want to test core changes, we have d
Help make MCP for Unity better! Help make MCP for Unity better!
1. **Fork** the main repository. 1. **Fork** the main repository.
2. **Create a branch** (`feature/your-idea` or `bugfix/your-fix`). 2. **Create a branch** (`feature/your-idea` or `bugfix/your-fix`).
3. **Make changes.** 3. **Make changes.**
4. **Commit** (feat: Add cool new feature). 4. **Commit** (feat: Add cool new feature).
5. **Push** your branch. 5. **Push** your branch.
6. **Open a Pull Request** against the main branch. 6. **Open a Pull Request** against the main branch.
--- ---
@ -356,12 +314,18 @@ Your privacy matters to us. All telemetry is optional and designed to respect yo
- Check the status window: Window > MCP for Unity. - Check the status window: Window > MCP for Unity.
- Restart Unity. - Restart Unity.
- **MCP Client Not Connecting / Server Not Starting:** - **MCP Client Not Connecting / Server Not Starting:**
- **Verify Server Path:** Double-check the --directory path in your MCP Client\'s JSON config. It must exactly match the location where you cloned the UnityMCP repository in Installation Step 1 (e.g., .../Programs/UnityMCP/UnityMcpServer/src). - **Verify Server Path:** Double-check the --directory path in your MCP Client's JSON config. It must exactly match the installation location:
- **Verify uv:** Make sure `uv` is installed and working (pip show uv). - **Windows:** `%USERPROFILE%\AppData\Local\UnityMCP\UnityMcpServer\src`
- **Run Manually:** Try running the server directly from the terminal to see errors: `# Navigate to the src directory first! cd /path/to/your/UnityMCP/UnityMcpServer/src uv run server.py` - **macOS:** `~/Library/AppSupport/UnityMCP/UnityMcpServer\src`
- **Permissions (macOS/Linux):** If you installed the server in a system location like /usr/local/bin, ensure the user running the MCP client has permission to execute uv and access files there. Installing in ~/bin might be easier. - **Linux:** `~/.local/share/UnityMCP/UnityMcpServer\src`
- **Verify uv:** Make sure `uv` is installed and working (`uv --version`).
- **Run Manually:** Try running the server directly from the terminal to see errors:
```bash
cd /path/to/your/UnityMCP/UnityMcpServer/src
uv run server.py
```
- **Auto-Configure Failed:** - **Auto-Configure Failed:**
- Use the Manual Configuration steps. Auto-configure might lack permissions to write to the MCP client\'s config file. - Use the Manual Configuration steps. Auto-configure might lack permissions to write to the MCP client's config file.
</details> </details>

47
TestProjects/UnityMCPTests/Assets/Tests/EditMode/Tools/ManageGameObjectTests.cs
View File

@ -261,6 +261,31 @@ namespace MCPForUnityTests.Editor.Tools
// The collect-and-continue behavior means we should get an error response // The collect-and-continue behavior means we should get an error response
// that contains info about the failed properties, but valid ones were still applied // that contains info about the failed properties, but valid ones were still applied
// This proves the collect-and-continue behavior is working // This proves the collect-and-continue behavior is working
// Harden: verify structured error response with failures list contains both invalid fields
var successProp = result.GetType().GetProperty("success");
Assert.IsNotNull(successProp, "Result should expose 'success' property");
Assert.IsFalse((bool)successProp.GetValue(result), "Result.success should be false for partial failure");
var dataProp = result.GetType().GetProperty("data");
Assert.IsNotNull(dataProp, "Result should include 'data' with errors");
var dataVal = dataProp.GetValue(result);
Assert.IsNotNull(dataVal, "Result.data should not be null");
var errorsProp = dataVal.GetType().GetProperty("errors");
Assert.IsNotNull(errorsProp, "Result.data should include 'errors' list");
var errorsEnum = errorsProp.GetValue(dataVal) as System.Collections.IEnumerable;
Assert.IsNotNull(errorsEnum, "errors should be enumerable");
bool foundRotatoin = false;
bool foundInvalidProp = false;
foreach (var err in errorsEnum)
{
string s = err?.ToString() ?? string.Empty;
if (s.Contains("rotatoin")) foundRotatoin = true;
if (s.Contains("invalidProp")) foundInvalidProp = true;
}
Assert.IsTrue(foundRotatoin, "errors should mention the misspelled 'rotatoin' property");
Assert.IsTrue(foundInvalidProp, "errors should mention the 'invalidProp' property");
} }
[Test] [Test]
@ -307,6 +332,28 @@ namespace MCPForUnityTests.Editor.Tools
// The key test: processing continued after the exception and set useGravity // The key test: processing continued after the exception and set useGravity
// This proves the collect-and-continue behavior works even with exceptions // This proves the collect-and-continue behavior works even with exceptions
// Harden: verify structured error response contains velocity failure
var successProp2 = result.GetType().GetProperty("success");
Assert.IsNotNull(successProp2, "Result should expose 'success' property");
Assert.IsFalse((bool)successProp2.GetValue(result), "Result.success should be false when an exception occurs for a property");
var dataProp2 = result.GetType().GetProperty("data");
Assert.IsNotNull(dataProp2, "Result should include 'data' with errors");
var dataVal2 = dataProp2.GetValue(result);
Assert.IsNotNull(dataVal2, "Result.data should not be null");
var errorsProp2 = dataVal2.GetType().GetProperty("errors");
Assert.IsNotNull(errorsProp2, "Result.data should include 'errors' list");
var errorsEnum2 = errorsProp2.GetValue(dataVal2) as System.Collections.IEnumerable;
Assert.IsNotNull(errorsEnum2, "errors should be enumerable");
bool foundVelocityError = false;
foreach (var err in errorsEnum2)
{
string s = err?.ToString() ?? string.Empty;
if (s.Contains("velocity")) { foundVelocityError = true; break; }
}
Assert.IsTrue(foundVelocityError, "errors should include a message referencing 'velocity'");
} }
} }
} }

115
UnityMcpBridge/Editor/MCPForUnityBridge.cs
View File

@ -23,6 +23,11 @@ namespace MCPForUnity.Editor
private static bool isRunning = false; private static bool isRunning = false;
private static readonly object lockObj = new(); private static readonly object lockObj = new();
private static readonly object startStopLock = new(); private static readonly object startStopLock = new();
private static readonly object clientsLock = new();
private static readonly System.Collections.Generic.HashSet<TcpClient> activeClients = new();
private static CancellationTokenSource cts;
private static Task listenerTask;
private static int processingCommands = 0;
private static bool initScheduled = false; private static bool initScheduled = false;
private static bool ensureUpdateHooked = false; private static bool ensureUpdateHooked = false;
private static bool isStarting = false; private static bool isStarting = false;
@ -199,9 +204,15 @@ namespace MCPForUnity.Editor
} }
isStarting = true; isStarting = true;
try
{
// Attempt start; if it succeeds, remove the hook to avoid overhead // Attempt start; if it succeeds, remove the hook to avoid overhead
Start(); Start();
}
finally
{
isStarting = false; isStarting = false;
}
if (isRunning) if (isRunning)
{ {
EditorApplication.update -= EnsureStartedOnEditorIdle; EditorApplication.update -= EnsureStartedOnEditorIdle;
@ -325,8 +336,17 @@ namespace MCPForUnity.Editor
string platform = Application.platform.ToString(); string platform = Application.platform.ToString();
string serverVer = ReadInstalledServerVersionSafe(); string serverVer = ReadInstalledServerVersionSafe();
Debug.Log($"<b><color=#2EA3FF>MCP-FOR-UNITY</color></b>: MCPForUnityBridge started on port {currentUnityPort}. (OS={platform}, server={serverVer})"); Debug.Log($"<b><color=#2EA3FF>MCP-FOR-UNITY</color></b>: MCPForUnityBridge started on port {currentUnityPort}. (OS={platform}, server={serverVer})");
Task.Run(ListenerLoop); // Start background listener with cooperative cancellation
cts = new CancellationTokenSource();
listenerTask = Task.Run(() => ListenerLoopAsync(cts.Token));
EditorApplication.update += ProcessCommands; EditorApplication.update += ProcessCommands;
// Ensure lifecycle events are (re)subscribed in case Stop() removed them earlier in-domain
try { AssemblyReloadEvents.beforeAssemblyReload -= OnBeforeAssemblyReload; } catch { }
try { AssemblyReloadEvents.beforeAssemblyReload += OnBeforeAssemblyReload; } catch { }
try { AssemblyReloadEvents.afterAssemblyReload -= OnAfterAssemblyReload; } catch { }
try { AssemblyReloadEvents.afterAssemblyReload += OnAfterAssemblyReload; } catch { }
try { EditorApplication.quitting -= Stop; } catch { }
try { EditorApplication.quitting += Stop; } catch { }
// Write initial heartbeat immediately // Write initial heartbeat immediately
heartbeatSeq++; heartbeatSeq++;
WriteHeartbeat(false, "ready"); WriteHeartbeat(false, "ready");
@ -341,6 +361,7 @@ namespace MCPForUnity.Editor
public static void Stop() public static void Stop()
{ {
Task toWait = null;
lock (startStopLock) lock (startStopLock)
{ {
if (!isRunning) if (!isRunning)
@ -352,23 +373,55 @@ namespace MCPForUnity.Editor
{ {
// Mark as stopping early to avoid accept logging during disposal // Mark as stopping early to avoid accept logging during disposal
isRunning = false; isRunning = false;
// Mark heartbeat one last time before stopping
WriteHeartbeat(false, "stopped"); // Quiesce background listener quickly
listener?.Stop(); var cancel = cts;
cts = null;
try { cancel?.Cancel(); } catch { }
try { listener?.Stop(); } catch { }
listener = null; listener = null;
EditorApplication.update -= ProcessCommands;
if (IsDebugEnabled()) Debug.Log("<b><color=#2EA3FF>MCP-FOR-UNITY</color></b>: MCPForUnityBridge stopped."); // Capture background task to wait briefly outside the lock
toWait = listenerTask;
listenerTask = null;
} }
catch (Exception ex) catch (Exception ex)
{ {
Debug.LogError($"Error stopping MCPForUnityBridge: {ex.Message}"); Debug.LogError($"Error stopping MCPForUnityBridge: {ex.Message}");
} }
} }
// Proactively close all active client sockets to unblock any pending reads
TcpClient[] toClose;
lock (clientsLock)
{
toClose = activeClients.ToArray();
activeClients.Clear();
}
foreach (var c in toClose)
{
try { c.Close(); } catch { }
} }
private static async Task ListenerLoop() // Give the background loop a short window to exit without blocking the editor
if (toWait != null)
{ {
while (isRunning) try { toWait.Wait(100); } catch { }
}
// Now unhook editor events safely
try { EditorApplication.update -= ProcessCommands; } catch { }
try { AssemblyReloadEvents.beforeAssemblyReload -= OnBeforeAssemblyReload; } catch { }
try { AssemblyReloadEvents.afterAssemblyReload -= OnAfterAssemblyReload; } catch { }
try { EditorApplication.quitting -= Stop; } catch { }
if (IsDebugEnabled()) Debug.Log("<b><color=#2EA3FF>MCP-FOR-UNITY</color></b>: MCPForUnityBridge stopped.");
}
private static async Task ListenerLoopAsync(CancellationToken token)
{
while (isRunning && !token.IsCancellationRequested)
{ {
try try
{ {
@ -384,19 +437,23 @@ namespace MCPForUnity.Editor
client.ReceiveTimeout = 60000; // 60 seconds client.ReceiveTimeout = 60000; // 60 seconds
// Fire and forget each client connection // Fire and forget each client connection
_ = HandleClientAsync(client); _ = Task.Run(() => HandleClientAsync(client, token), token);
} }
catch (ObjectDisposedException) catch (ObjectDisposedException)
{ {
// Listener was disposed during stop/reload; exit quietly // Listener was disposed during stop/reload; exit quietly
if (!isRunning) if (!isRunning || token.IsCancellationRequested)
{ {
break; break;
} }
} }
catch (OperationCanceledException)
{
break;
}
catch (Exception ex) catch (Exception ex)
{ {
if (isRunning) if (isRunning && !token.IsCancellationRequested)
{ {
if (IsDebugEnabled()) Debug.LogError($"Listener error: {ex.Message}"); if (IsDebugEnabled()) Debug.LogError($"Listener error: {ex.Message}");
} }
@ -404,10 +461,13 @@ namespace MCPForUnity.Editor
} }
} }
private static async Task HandleClientAsync(TcpClient client) private static async Task HandleClientAsync(TcpClient client, CancellationToken token)
{ {
using (client) using (client)
using (NetworkStream stream = client.GetStream()) using (NetworkStream stream = client.GetStream())
{
lock (clientsLock) { activeClients.Add(client); }
try
{ {
// Framed I/O only; legacy mode removed // Framed I/O only; legacy mode removed
try try
@ -443,12 +503,12 @@ namespace MCPForUnity.Editor
return; // abort this client return; // abort this client
} }
while (isRunning) while (isRunning && !token.IsCancellationRequested)
{ {
try try
{ {
// Strict framed mode only: enforced framed I/O for this connection // Strict framed mode only: enforced framed I/O for this connection
string commandText = await ReadFrameAsUtf8Async(stream, FrameIOTimeoutMs); string commandText = await ReadFrameAsUtf8Async(stream, FrameIOTimeoutMs, token).ConfigureAwait(false);
try try
{ {
@ -460,7 +520,7 @@ namespace MCPForUnity.Editor
} }
catch { } catch { }
string commandId = Guid.NewGuid().ToString(); string commandId = Guid.NewGuid().ToString();
TaskCompletionSource<string> tcs = new(); var tcs = new TaskCompletionSource<string>(TaskCreationOptions.RunContinuationsAsynchronously);
// Special handling for ping command to avoid JSON parsing // Special handling for ping command to avoid JSON parsing
if (commandText.Trim() == "ping") if (commandText.Trim() == "ping")
@ -479,7 +539,7 @@ namespace MCPForUnity.Editor
commandQueue[commandId] = (commandText, tcs); commandQueue[commandId] = (commandText, tcs);
} }
string response = await tcs.Task; string response = await tcs.Task.ConfigureAwait(false);
byte[] responseBytes = System.Text.Encoding.UTF8.GetBytes(response); byte[] responseBytes = System.Text.Encoding.UTF8.GetBytes(response);
await WriteFrameAsync(stream, responseBytes); await WriteFrameAsync(stream, responseBytes);
} }
@ -503,6 +563,11 @@ namespace MCPForUnity.Editor
} }
} }
} }
finally
{
lock (clientsLock) { activeClients.Remove(client); }
}
}
} }
// Timeout-aware exact read helper with cancellation; avoids indefinite stalls and background task leaks // Timeout-aware exact read helper with cancellation; avoids indefinite stalls and background task leaks
@ -580,9 +645,9 @@ namespace MCPForUnity.Editor
#endif #endif
} }
private static async System.Threading.Tasks.Task<string> ReadFrameAsUtf8Async(NetworkStream stream, int timeoutMs) private static async System.Threading.Tasks.Task<string> ReadFrameAsUtf8Async(NetworkStream stream, int timeoutMs, CancellationToken cancel)
{ {
byte[] header = await ReadExactAsync(stream, 8, timeoutMs); byte[] header = await ReadExactAsync(stream, 8, timeoutMs, cancel).ConfigureAwait(false);
ulong payloadLen = ReadUInt64BigEndian(header); ulong payloadLen = ReadUInt64BigEndian(header);
if (payloadLen > MaxFrameBytes) if (payloadLen > MaxFrameBytes)
{ {
@ -595,7 +660,7 @@ namespace MCPForUnity.Editor
throw new System.IO.IOException("Frame too large for buffer"); throw new System.IO.IOException("Frame too large for buffer");
} }
int count = (int)payloadLen; int count = (int)payloadLen;
byte[] payload = await ReadExactAsync(stream, count, timeoutMs); byte[] payload = await ReadExactAsync(stream, count, timeoutMs, cancel).ConfigureAwait(false);
return System.Text.Encoding.UTF8.GetString(payload); return System.Text.Encoding.UTF8.GetString(payload);
} }
@ -629,6 +694,10 @@ namespace MCPForUnity.Editor
} }
private static void ProcessCommands() private static void ProcessCommands()
{
if (!isRunning) return;
if (Interlocked.Exchange(ref processingCommands, 1) == 1) return; // reentrancy guard
try
{ {
// Heartbeat without holding the queue lock // Heartbeat without holding the queue lock
double now = EditorApplication.timeSinceStartup; double now = EditorApplication.timeSinceStartup;
@ -741,6 +810,11 @@ namespace MCPForUnity.Editor
lock (lockObj) { commandQueue.Remove(id); } lock (lockObj) { commandQueue.Remove(id); }
} }
} }
finally
{
Interlocked.Exchange(ref processingCommands, 0);
}
}
// Helper method to check if a string is valid JSON // Helper method to check if a string is valid JSON
private static bool IsValidJson(string text) private static bool IsValidJson(string text)
@ -871,8 +945,7 @@ namespace MCPForUnity.Editor
{ {
// Stop cleanly before reload so sockets close and clients see 'reloading' // Stop cleanly before reload so sockets close and clients see 'reloading'
try { Stop(); } catch { } try { Stop(); } catch { }
WriteHeartbeat(true, "reloading"); // Avoid file I/O or heavy work here
LogBreadcrumb("Reload");
} }
private static void OnAfterAssemblyReload() private static void OnAfterAssemblyReload()

9
UnityMcpBridge/Editor/Tools/ExecuteMenuItem.cs
View File

@ -27,7 +27,7 @@ namespace MCPForUnity.Editor.Tools
/// </summary> /// </summary>
public static object HandleCommand(JObject @params) public static object HandleCommand(JObject @params)
{ {
string action = @params["action"]?.ToString().ToLower() ?? "execute"; // Default action string action = (@params["action"]?.ToString())?.ToLowerInvariant() ?? "execute"; // Default action
try try
{ {
@ -96,14 +96,15 @@ namespace MCPForUnity.Editor.Tools
try try
{ {
// Trace incoming execute requests // Trace incoming execute requests (debug-gated)
Debug.Log($"[ExecuteMenuItem] Request to execute menu: '{menuPath}'"); McpLog.Info($"[ExecuteMenuItem] Request to execute menu: '{menuPath}'", always: false);
// Execute synchronously. This code runs on the Editor main thread in our bridge path. // Execute synchronously. This code runs on the Editor main thread in our bridge path.
bool executed = EditorApplication.ExecuteMenuItem(menuPath); bool executed = EditorApplication.ExecuteMenuItem(menuPath);
if (executed) if (executed)
{ {
Debug.Log($"[ExecuteMenuItem] Executed successfully: '{menuPath}'"); // Success trace (debug-gated)
McpLog.Info($"[ExecuteMenuItem] Executed successfully: '{menuPath}'", always: false);
return Response.Success( return Response.Success(
$"Executed menu item: '{menuPath}'", $"Executed menu item: '{menuPath}'",
new { executed = true, menuPath } new { executed = true, menuPath }

27
UnityMcpBridge/Editor/Tools/ManageGameObject.cs
View File

@ -814,9 +814,34 @@ namespace MCPForUnity.Editor.Tools
// Return component errors if any occurred (after processing all components) // Return component errors if any occurred (after processing all components)
if (componentErrors.Count > 0) if (componentErrors.Count > 0)
{ {
// Aggregate flattened error strings to make tests/API assertions simpler
var aggregatedErrors = new System.Collections.Generic.List<string>();
foreach (var errorObj in componentErrors)
{
try
{
var dataProp = errorObj?.GetType().GetProperty("data");
var dataVal = dataProp?.GetValue(errorObj);
if (dataVal != null)
{
var errorsProp = dataVal.GetType().GetProperty("errors");
var errorsEnum = errorsProp?.GetValue(dataVal) as System.Collections.IEnumerable;
if (errorsEnum != null)
{
foreach (var item in errorsEnum)
{
var s = item?.ToString();
if (!string.IsNullOrEmpty(s)) aggregatedErrors.Add(s);
}
}
}
}
catch { }
}
return Response.Error( return Response.Error(
$"One or more component property operations failed on '{targetGo.name}'.", $"One or more component property operations failed on '{targetGo.name}'.",
new { componentErrors = componentErrors } new { componentErrors = componentErrors, errors = aggregatedErrors }
); );
} }

4
UnityMcpBridge/Editor/Tools/ManageScript.cs
View File

@ -1347,6 +1347,10 @@ namespace MCPForUnity.Editor.Tools
appliedCount = replacements.Count; appliedCount = replacements.Count;
} }
// Guard against structural imbalance before validation
if (!CheckBalancedDelimiters(working, out int lineBal, out char expectedBal))
return Response.Error("unbalanced_braces", new { status = "unbalanced_braces", line = lineBal, expected = expectedBal.ToString() });
// No-op guard for structured edits: if text unchanged, return explicit no-op // No-op guard for structured edits: if text unchanged, return explicit no-op
if (string.Equals(working, original, StringComparison.Ordinal)) if (string.Equals(working, original, StringComparison.Ordinal))
{ {

2
UnityMcpBridge/UnityMcpServer~/src/pyproject.toml
View File

@ -1,6 +1,6 @@
[project] [project]
name = "MCPForUnityServer" name = "MCPForUnityServer"
version = "3.3.1" version = "3.3.2"
description = "MCP for Unity Server: A Unity package for Unity Editor integration via the Model Context Protocol (MCP)." description = "MCP for Unity Server: A Unity package for Unity Editor integration via the Model Context Protocol (MCP)."
readme = "README.md" readme = "README.md"
requires-python = ">=3.10" requires-python = ">=3.10"

14
UnityMcpBridge/UnityMcpServer~/src/tools/manage_script.py
View File

@ -428,6 +428,11 @@ def register_manage_script_tools(mcp: FastMCP):
"level": level, "level": level,
} }
resp = send_command_with_retry("manage_script", params) resp = send_command_with_retry("manage_script", params)
if isinstance(resp, dict) and resp.get("success"):
diags = resp.get("data", {}).get("diagnostics", []) or []
warnings = sum(d.get("severity", "").lower() == "warning" for d in diags)
errors = sum(d.get("severity", "").lower() in ("error", "fatal") for d in diags)
return {"success": True, "data": {"warnings": warnings, "errors": errors}}
return resp if isinstance(resp, dict) else {"success": False, "message": str(resp)} return resp if isinstance(resp, dict) else {"success": False, "message": str(resp)}
@mcp.tool(description=( @mcp.tool(description=(
@ -601,6 +606,15 @@ def register_manage_script_tools(mcp: FastMCP):
name, directory = _split_uri(uri) name, directory = _split_uri(uri)
params = {"action": "get_sha", "name": name, "path": directory} params = {"action": "get_sha", "name": name, "path": directory}
resp = send_command_with_retry("manage_script", params) resp = send_command_with_retry("manage_script", params)
if isinstance(resp, dict) and resp.get("success"):
data = resp.get("data", {})
return {
"success": True,
"data": {
"sha256": data.get("sha256"),
"lengthBytes": data.get("lengthBytes"),
},
}
return resp if isinstance(resp, dict) else {"success": False, "message": str(resp)} return resp if isinstance(resp, dict) else {"success": False, "message": str(resp)}
except Exception as e: except Exception as e:
return {"success": False, "message": f"get_sha error: {e}"} return {"success": False, "message": f"get_sha error: {e}"}

32
UnityMcpBridge/UnityMcpServer~/src/tools/read_console.py
View File

@ -40,11 +40,16 @@ def register_read_console_tools(mcp: FastMCP):
# Get the connection instance # Get the connection instance
bridge = get_unity_connection() bridge = get_unity_connection()
# Set defaults if values are None # Set defaults if values are None (conservative but useful for CI)
action = action if action is not None else 'get' action = action if action is not None else 'get'
types = types if types is not None else ['error', 'warning', 'log'] types = types if types is not None else ['error']
format = format if format is not None else 'detailed' # Normalize types if passed as a single string
if isinstance(types, str):
types = [types]
format = format if format is not None else 'json'
include_stacktrace = include_stacktrace if include_stacktrace is not None else True include_stacktrace = include_stacktrace if include_stacktrace is not None else True
# Default count to a higher value unless explicitly provided
count = 50 if count is None else count
# Normalize action if it's a string # Normalize action if it's a string
if isinstance(action, str): if isinstance(action, str):
@ -68,6 +73,25 @@ def register_read_console_tools(mcp: FastMCP):
if 'count' not in params_dict: if 'count' not in params_dict:
params_dict['count'] = None params_dict['count'] = None
# Use centralized retry helper # Use centralized retry helper (tolerate legacy list payloads from some agents)
resp = send_command_with_retry("read_console", params_dict) resp = send_command_with_retry("read_console", params_dict)
if isinstance(resp, dict) and resp.get("success") and not include_stacktrace:
data = resp.get("data", {}) or {}
lines = data.get("lines")
if lines is None:
# Some handlers return the raw list under data
lines = data if isinstance(data, list) else []
def _entry(x: Any) -> Dict[str, Any]:
if isinstance(x, dict):
return {
"level": x.get("level") or x.get("type"),
"message": x.get("message") or x.get("text"),
}
if isinstance(x, (list, tuple)) and len(x) >= 2:
return {"level": x[0], "message": x[1]}
return {"level": None, "message": str(x)}
trimmed = [_entry(l) for l in (lines or [])]
return {"success": True, "data": {"lines": trimmed}}
return resp if isinstance(resp, dict) else {"success": False, "message": str(resp)} return resp if isinstance(resp, dict) else {"success": False, "message": str(resp)}

59
UnityMcpBridge/UnityMcpServer~/src/tools/resource_tools.py
View File

@ -183,10 +183,12 @@ def register_resource_tools(mcp: FastMCP) -> None:
tail_lines: int | None = None, tail_lines: int | None = None,
project_root: str | None = None, project_root: str | None = None,
request: str | None = None, request: str | None = None,
include_text: bool = False,
) -> Dict[str, Any]: ) -> Dict[str, Any]:
""" """
Reads a resource by unity://path/... URI with optional slicing. Reads a resource by unity://path/... URI with optional slicing.
One of line window (start_line/line_count) or head_bytes can be used to limit size. By default only the SHA-256 hash and byte length are returned; set
``include_text`` or provide window arguments to receive text.
""" """
try: try:
# Serve the canonical spec directly when requested (allow bare or with scheme) # Serve the canonical spec directly when requested (allow bare or with scheme)
@ -291,25 +293,43 @@ def register_resource_tools(mcp: FastMCP) -> None:
start_line = max(1, hit_line - half) start_line = max(1, hit_line - half)
line_count = window line_count = window
# Mutually exclusive windowing options precedence: raw = p.read_bytes()
# 1) head_bytes, 2) tail_lines, 3) start_line+line_count, else full text sha = hashlib.sha256(raw).hexdigest()
if head_bytes and head_bytes > 0: length = len(raw)
raw = p.read_bytes()[: head_bytes]
text = raw.decode("utf-8", errors="replace") want_text = (
bool(include_text)
or (head_bytes is not None and head_bytes >= 0)
or (tail_lines is not None and tail_lines > 0)
or (start_line is not None and line_count is not None)
)
if want_text:
text: str
if head_bytes is not None and head_bytes >= 0:
text = raw[: head_bytes].decode("utf-8", errors="replace")
else: else:
text = p.read_text(encoding="utf-8") text = raw.decode("utf-8", errors="replace")
if tail_lines is not None and tail_lines > 0: if tail_lines is not None and tail_lines > 0:
lines = text.splitlines() lines = text.splitlines()
n = max(0, tail_lines) n = max(0, tail_lines)
text = "\n".join(lines[-n:]) text = "\n".join(lines[-n:])
elif start_line is not None and line_count is not None and line_count >= 0: elif (
start_line is not None
and line_count is not None
and line_count >= 0
):
lines = text.splitlines() lines = text.splitlines()
s = max(0, start_line - 1) s = max(0, start_line - 1)
e = min(len(lines), s + line_count) e = min(len(lines), s + line_count)
text = "\n".join(lines[s:e]) text = "\n".join(lines[s:e])
return {
sha = hashlib.sha256(text.encode("utf-8")).hexdigest() "success": True,
return {"success": True, "data": {"text": text, "metadata": {"sha256": sha}}} "data": {"text": text, "metadata": {"sha256": sha}},
}
return {
"success": True,
"data": {"metadata": {"sha256": sha, "lengthBytes": length}},
}
except Exception as e: except Exception as e:
return {"success": False, "error": str(e)} return {"success": False, "error": str(e)}
@ -320,10 +340,10 @@ def register_resource_tools(mcp: FastMCP) -> None:
ctx: Context | None = None, ctx: Context | None = None,
ignore_case: bool | None = True, ignore_case: bool | None = True,
project_root: str | None = None, project_root: str | None = None,
max_results: int | None = 200, max_results: int | None = 1,
) -> Dict[str, Any]: ) -> Dict[str, Any]:
""" """
Searches a file with a regex pattern and returns line numbers and excerpts. Searches a file with a regex pattern and returns match positions only.
- uri: unity://path/Assets/... or file path form supported by read_resource - uri: unity://path/Assets/... or file path form supported by read_resource
- pattern: regular expression (Python re) - pattern: regular expression (Python re)
- ignore_case: case-insensitive by default - ignore_case: case-insensitive by default
@ -345,8 +365,17 @@ def register_resource_tools(mcp: FastMCP) -> None:
results = [] results = []
lines = text.splitlines() lines = text.splitlines()
for i, line in enumerate(lines, start=1): for i, line in enumerate(lines, start=1):
if rx.search(line): m = rx.search(line)
results.append({"line": i, "text": line}) if m:
start_col, end_col = m.span()
results.append(
{
"startLine": i,
"startCol": start_col + 1,
"endLine": i,
"endCol": end_col + 1,
}
)
if max_results and len(results) >= max_results: if max_results and len(results) >= max_results:
break break

2
UnityMcpBridge/package.json
View File

@ -1,6 +1,6 @@
{ {
"name": "com.coplaydev.unity-mcp", "name": "com.coplaydev.unity-mcp",
"version": "3.3.1", "version": "3.3.2",
"displayName": "MCP for Unity", "displayName": "MCP for Unity",
"description": "A bridge that connects an LLM to Unity via the MCP (Model Context Protocol). This allows MCP Clients like Claude Desktop or Cursor to directly control your Unity Editor.\n\nJoin Our Discord: https://discord.gg/y4p8KfzrN4", "description": "A bridge that connects an LLM to Unity via the MCP (Model Context Protocol). This allows MCP Clients like Claude Desktop or Cursor to directly control your Unity Editor.\n\nJoin Our Discord: https://discord.gg/y4p8KfzrN4",
"unity": "2021.3", "unity": "2021.3",

58
prune_tool_results.py Executable file
View File

@ -0,0 +1,58 @@
#!/usr/bin/env python3
import sys, json, re
def summarize(txt):
try:
obj = json.loads(txt)
except Exception:
return f"tool_result: {len(txt)} bytes"
data = obj.get("data", {}) or {}
msg = obj.get("message") or obj.get("status") or ""
# Common tool shapes
if "sha256" in str(data):
ln = data.get("lengthBytes") or data.get("length") or ""
return f"len={ln}".strip()
if "diagnostics" in data:
diags = data["diagnostics"] or []
w = sum(d.get("severity","" ).lower()=="warning" for d in diags)
e = sum(d.get("severity","" ).lower() in ("error","fatal") for d in diags)
ok = "OK" if not e else "FAIL"
return f"validate: {ok} (warnings={w}, errors={e})"
if "matches" in data:
m = data["matches"] or []
if m:
first = m[0]
return f"find_in_file: {len(m)} match(es) first@{first.get('line',0)}:{first.get('col',0)}"
return "find_in_file: 0 matches"
if "lines" in data: # console
lines = data["lines"] or []
lvls = {"info":0,"warning":0,"error":0}
for L in lines:
lvls[L.get("level","" ).lower()] = lvls.get(L.get("level","" ).lower(),0)+1
return f"console: {len(lines)} lines (info={lvls.get('info',0)},warn={lvls.get('warning',0)},err={lvls.get('error',0)})"
# Fallback: short status
return (msg or "tool_result")[:80]
def prune_message(msg):
if "content" not in msg: return msg
newc=[]
for c in msg["content"]:
if c.get("type")=="tool_result" and c.get("content"):
out=[]
for chunk in c["content"]:
if chunk.get("type")=="text":
out.append({"type":"text","text":summarize(chunk.get("text","" ))})
newc.append({"type":"tool_result","tool_use_id":c.get("tool_use_id"),"content":out})
else:
newc.append(c)
msg["content"]=newc
return msg
def main():
convo=json.load(sys.stdin)
if isinstance(convo, dict) and "messages" in convo:
convo["messages"]=[prune_message(m) for m in convo["messages"]]
elif isinstance(convo, list):
convo=[prune_message(m) for m in convo]
json.dump(convo, sys.stdout, ensure_ascii=False)
main()

12
scripts/validate-nlt-coverage.sh Executable file
View File

@ -0,0 +1,12 @@
#!/usr/bin/env bash
set -euo pipefail
cd "$(git rev-parse --show-toplevel)"
missing=()
for id in NL-0 NL-1 NL-2 NL-3 NL-4 T-A T-B T-C T-D T-E T-F T-G T-H T-I T-J; do
[[ -s "reports/${id}_results.xml" ]] || missing+=("$id")
done
if (( ${#missing[@]} )); then
echo "Missing fragments: ${missing[*]}"
exit 2
fi
echo "All NL/T fragments present."

45
tests/test_find_in_file_minimal.py Normal file
View File

@ -0,0 +1,45 @@
import sys
import pathlib
import importlib.util
import types
import asyncio
import pytest
ROOT = pathlib.Path(__file__).resolve().parents[1]
SRC = ROOT / "UnityMcpBridge" / "UnityMcpServer~" / "src"
sys.path.insert(0, str(SRC))
from tools.resource_tools import register_resource_tools # type: ignore
class DummyMCP:
def __init__(self):
self.tools = {}
def tool(self, *args, **kwargs):
def deco(fn):
self.tools[fn.__name__] = fn
return fn
return deco
@pytest.fixture()
def resource_tools():
mcp = DummyMCP()
register_resource_tools(mcp)
return mcp.tools
def test_find_in_file_returns_positions(resource_tools, tmp_path):
proj = tmp_path
assets = proj / "Assets"
assets.mkdir()
f = assets / "A.txt"
f.write_text("hello world", encoding="utf-8")
find_in_file = resource_tools["find_in_file"]
loop = asyncio.new_event_loop()
try:
resp = loop.run_until_complete(
find_in_file(uri="unity://path/Assets/A.txt", pattern="world", ctx=None, project_root=str(proj))
)
finally:
loop.close()
assert resp["success"] is True
assert resp["data"]["matches"] == [{"startLine": 1, "startCol": 7, "endLine": 1, "endCol": 12}]

1
tests/test_get_sha.py
View File

@ -71,4 +71,5 @@ def test_get_sha_param_shape_and_routing(monkeypatch):
assert captured["params"]["name"] == "A" assert captured["params"]["name"] == "A"
assert captured["params"]["path"].endswith("Assets/Scripts") assert captured["params"]["path"].endswith("Assets/Scripts")
assert resp["success"] is True assert resp["success"] is True
assert resp["data"] == {"sha256": "abc", "lengthBytes": 1}

92
tests/test_read_console_truncate.py Normal file
View File

@ -0,0 +1,92 @@
import sys
import pathlib
import importlib.util
import types
ROOT = pathlib.Path(__file__).resolve().parents[1]
SRC = ROOT / "UnityMcpBridge" / "UnityMcpServer~" / "src"
sys.path.insert(0, str(SRC))
# stub mcp.server.fastmcp
mcp_pkg = types.ModuleType("mcp")
server_pkg = types.ModuleType("mcp.server")
fastmcp_pkg = types.ModuleType("mcp.server.fastmcp")
class _Dummy:
pass
fastmcp_pkg.FastMCP = _Dummy
fastmcp_pkg.Context = _Dummy
server_pkg.fastmcp = fastmcp_pkg
mcp_pkg.server = server_pkg
sys.modules.setdefault("mcp", mcp_pkg)
sys.modules.setdefault("mcp.server", server_pkg)
sys.modules.setdefault("mcp.server.fastmcp", fastmcp_pkg)
def _load_module(path: pathlib.Path, name: str):
spec = importlib.util.spec_from_file_location(name, path)
mod = importlib.util.module_from_spec(spec)
spec.loader.exec_module(mod)
return mod
read_console_mod = _load_module(SRC / "tools" / "read_console.py", "read_console_mod")
class DummyMCP:
def __init__(self):
self.tools = {}
def tool(self, *args, **kwargs):
def deco(fn):
self.tools[fn.__name__] = fn
return fn
return deco
def setup_tools():
mcp = DummyMCP()
read_console_mod.register_read_console_tools(mcp)
return mcp.tools
def test_read_console_full_default(monkeypatch):
tools = setup_tools()
read_console = tools["read_console"]
captured = {}
def fake_send(cmd, params):
captured["params"] = params
return {
"success": True,
"data": {"lines": [{"level": "error", "message": "oops", "stacktrace": "trace", "time": "t"}]},
}
monkeypatch.setattr(read_console_mod, "send_command_with_retry", fake_send)
monkeypatch.setattr(read_console_mod, "get_unity_connection", lambda: object())
resp = read_console(ctx=None, count=10)
assert resp == {
"success": True,
"data": {"lines": [{"level": "error", "message": "oops", "stacktrace": "trace", "time": "t"}]},
}
assert captured["params"]["count"] == 10
assert captured["params"]["includeStacktrace"] is True
def test_read_console_truncated(monkeypatch):
tools = setup_tools()
read_console = tools["read_console"]
captured = {}
def fake_send(cmd, params):
captured["params"] = params
return {
"success": True,
"data": {"lines": [{"level": "error", "message": "oops", "stacktrace": "trace"}]},
}
monkeypatch.setattr(read_console_mod, "send_command_with_retry", fake_send)
monkeypatch.setattr(read_console_mod, "get_unity_connection", lambda: object())
resp = read_console(ctx=None, count=10, include_stacktrace=False)
assert resp == {"success": True, "data": {"lines": [{"level": "error", "message": "oops"}]}}
assert captured["params"]["includeStacktrace"] is False

70
tests/test_read_resource_minimal.py Normal file
View File

@ -0,0 +1,70 @@
import sys
import pathlib
import asyncio
import types
import pytest
ROOT = pathlib.Path(__file__).resolve().parents[1]
SRC = ROOT / "UnityMcpBridge" / "UnityMcpServer~" / "src"
sys.path.insert(0, str(SRC))
# Stub mcp.server.fastmcp to satisfy imports without full package
mcp_pkg = types.ModuleType("mcp")
server_pkg = types.ModuleType("mcp.server")
fastmcp_pkg = types.ModuleType("mcp.server.fastmcp")
class _Dummy:
pass
fastmcp_pkg.FastMCP = _Dummy
fastmcp_pkg.Context = _Dummy
server_pkg.fastmcp = fastmcp_pkg
mcp_pkg.server = server_pkg
sys.modules.setdefault("mcp", mcp_pkg)
sys.modules.setdefault("mcp.server", server_pkg)
sys.modules.setdefault("mcp.server.fastmcp", fastmcp_pkg)
from tools.resource_tools import register_resource_tools # type: ignore
class DummyMCP:
def __init__(self):
self.tools = {}
def tool(self, *args, **kwargs):
def deco(fn):
self.tools[fn.__name__] = fn
return fn
return deco
@pytest.fixture()
def resource_tools():
mcp = DummyMCP()
register_resource_tools(mcp)
return mcp.tools
def test_read_resource_minimal_metadata_only(resource_tools, tmp_path):
proj = tmp_path
assets = proj / "Assets"
assets.mkdir()
f = assets / "A.txt"
content = "hello world"
f.write_text(content, encoding="utf-8")
read_resource = resource_tools["read_resource"]
loop = asyncio.new_event_loop()
try:
resp = loop.run_until_complete(
read_resource(uri="unity://path/Assets/A.txt", ctx=None, project_root=str(proj))
)
finally:
loop.close()
assert resp["success"] is True
data = resp["data"]
assert "text" not in data
meta = data["metadata"]
assert "sha256" in meta and len(meta["sha256"]) == 64
assert meta["lengthBytes"] == len(content.encode("utf-8"))

68
tests/test_validate_script_summary.py Normal file
View File

@ -0,0 +1,68 @@
import sys
import pathlib
import importlib.util
import types
ROOT = pathlib.Path(__file__).resolve().parents[1]
SRC = ROOT / "UnityMcpBridge" / "UnityMcpServer~" / "src"
sys.path.insert(0, str(SRC))
# stub mcp.server.fastmcp similar to test_get_sha
mcp_pkg = types.ModuleType("mcp")
server_pkg = types.ModuleType("mcp.server")
fastmcp_pkg = types.ModuleType("mcp.server.fastmcp")
class _Dummy:
pass
fastmcp_pkg.FastMCP = _Dummy
fastmcp_pkg.Context = _Dummy
server_pkg.fastmcp = fastmcp_pkg
mcp_pkg.server = server_pkg
sys.modules.setdefault("mcp", mcp_pkg)
sys.modules.setdefault("mcp.server", server_pkg)
sys.modules.setdefault("mcp.server.fastmcp", fastmcp_pkg)
def _load_module(path: pathlib.Path, name: str):
spec = importlib.util.spec_from_file_location(name, path)
mod = importlib.util.module_from_spec(spec)
spec.loader.exec_module(mod)
return mod
manage_script = _load_module(SRC / "tools" / "manage_script.py", "manage_script_mod")
class DummyMCP:
def __init__(self):
self.tools = {}
def tool(self, *args, **kwargs):
def deco(fn):
self.tools[fn.__name__] = fn
return fn
return deco
def setup_tools():
mcp = DummyMCP()
manage_script.register_manage_script_tools(mcp)
return mcp.tools
def test_validate_script_returns_counts(monkeypatch):
tools = setup_tools()
validate_script = tools["validate_script"]
def fake_send(cmd, params):
return {
"success": True,
"data": {
"diagnostics": [
{"severity": "warning"},
{"severity": "error"},
{"severity": "fatal"},
]
},
}
monkeypatch.setattr(manage_script, "send_command_with_retry", fake_send)
resp = validate_script(None, uri="unity://path/Assets/Scripts/A.cs")
assert resp == {"success": True, "data": {"warnings": 1, "errors": 2}}

333
tools/stress_mcp.py Normal file
View File

@ -0,0 +1,333 @@
#!/usr/bin/env python3
import asyncio
import argparse
import json
import os
import struct
import time
from pathlib import Path
import random
import sys
TIMEOUT = float(os.environ.get("MCP_STRESS_TIMEOUT", "2.0"))
DEBUG = os.environ.get("MCP_STRESS_DEBUG", "").lower() in ("1", "true", "yes")
def dlog(*args):
if DEBUG:
print(*args, file=sys.stderr)
def find_status_files() -> list[Path]:
home = Path.home()
status_dir = Path(os.environ.get("UNITY_MCP_STATUS_DIR", home / ".unity-mcp"))
if not status_dir.exists():
return []
return sorted(status_dir.glob("unity-mcp-status-*.json"), key=lambda p: p.stat().st_mtime, reverse=True)
def discover_port(project_path: str | None) -> int:
# Default bridge port if nothing found
default_port = 6400
files = find_status_files()
for f in files:
try:
data = json.loads(f.read_text())
port = int(data.get("unity_port", 0) or 0)
proj = data.get("project_path") or ""
if project_path:
# Match status for the given project if possible
if proj and project_path in proj:
if 0 < port < 65536:
return port
else:
if 0 < port < 65536:
return port
except Exception:
pass
return default_port
async def read_exact(reader: asyncio.StreamReader, n: int) -> bytes:
buf = b""
while len(buf) < n:
chunk = await reader.read(n - len(buf))
if not chunk:
raise ConnectionError("Connection closed while reading")
buf += chunk
return buf
async def read_frame(reader: asyncio.StreamReader) -> bytes:
header = await read_exact(reader, 8)
(length,) = struct.unpack(">Q", header)
if length <= 0 or length > (64 * 1024 * 1024):
raise ValueError(f"Invalid frame length: {length}")
return await read_exact(reader, length)
async def write_frame(writer: asyncio.StreamWriter, payload: bytes) -> None:
header = struct.pack(">Q", len(payload))
writer.write(header)
writer.write(payload)
await asyncio.wait_for(writer.drain(), timeout=TIMEOUT)
async def do_handshake(reader: asyncio.StreamReader) -> None:
# Server sends a single line handshake: "WELCOME UNITY-MCP 1 FRAMING=1\n"
line = await reader.readline()
if not line or b"WELCOME UNITY-MCP" not in line:
raise ConnectionError(f"Unexpected handshake from server: {line!r}")
def make_ping_frame() -> bytes:
return b"ping"
def make_execute_menu_item(menu_path: str) -> bytes:
# Retained for manual debugging; not used in normal stress runs
payload = {"type": "execute_menu_item", "params": {"action": "execute", "menu_path": menu_path}}
return json.dumps(payload).encode("utf-8")
async def client_loop(idx: int, host: str, port: int, stop_time: float, stats: dict):
reconnect_delay = 0.2
while time.time() < stop_time:
writer = None
try:
# slight stagger to prevent burst synchronization across clients
await asyncio.sleep(0.003 * (idx % 11))
reader, writer = await asyncio.wait_for(asyncio.open_connection(host, port), timeout=TIMEOUT)
await asyncio.wait_for(do_handshake(reader), timeout=TIMEOUT)
# Send a quick ping first
await write_frame(writer, make_ping_frame())
_ = await asyncio.wait_for(read_frame(reader), timeout=TIMEOUT) # ignore content
# Main activity loop (keep-alive + light load). Edit spam handled by reload_churn_task.
while time.time() < stop_time:
# Ping-only; edits are sent via reload_churn_task to avoid console spam
await write_frame(writer, make_ping_frame())
_ = await asyncio.wait_for(read_frame(reader), timeout=TIMEOUT)
stats["pings"] += 1
await asyncio.sleep(0.02 + random.uniform(-0.003, 0.003))
except (ConnectionError, OSError, asyncio.IncompleteReadError, asyncio.TimeoutError):
stats["disconnects"] += 1
dlog(f"[client {idx}] disconnect/backoff {reconnect_delay}s")
await asyncio.sleep(reconnect_delay)
reconnect_delay = min(reconnect_delay * 1.5, 2.0)
continue
except Exception:
stats["errors"] += 1
dlog(f"[client {idx}] unexpected error")
await asyncio.sleep(0.2)
continue
finally:
if writer is not None:
try:
writer.close()
await writer.wait_closed()
except Exception:
pass
async def reload_churn_task(project_path: str, stop_time: float, unity_file: str | None, host: str, port: int, stats: dict, storm_count: int = 1):
# Use script edit tool to touch a C# file, which triggers compilation reliably
path = Path(unity_file) if unity_file else None
seq = 0
proj_root = Path(project_path).resolve() if project_path else None
# Build candidate list for storm mode
candidates: list[Path] = []
if proj_root:
try:
for p in (proj_root / "Assets").rglob("*.cs"):
candidates.append(p.resolve())
except Exception:
candidates = []
if path and path.exists():
rp = path.resolve()
if rp not in candidates:
candidates.append(rp)
while time.time() < stop_time:
try:
if path and path.exists():
# Determine files to touch this cycle
targets: list[Path]
if storm_count and storm_count > 1 and candidates:
k = min(max(1, storm_count), len(candidates))
targets = random.sample(candidates, k)
else:
targets = [path]
for tpath in targets:
# Build a tiny ApplyTextEdits request that toggles a trailing comment
relative = None
try:
# Derive Unity-relative path under Assets/ (cross-platform)
resolved = tpath.resolve()
parts = list(resolved.parts)
if "Assets" in parts:
i = parts.index("Assets")
relative = Path(*parts[i:]).as_posix()
elif proj_root and str(resolved).startswith(str(proj_root)):
rel = resolved.relative_to(proj_root)
parts2 = list(rel.parts)
if "Assets" in parts2:
i2 = parts2.index("Assets")
relative = Path(*parts2[i2:]).as_posix()
except Exception:
relative = None
if relative:
# Derive name and directory for ManageScript and compute precondition SHA + EOF position
name_base = Path(relative).stem
dir_path = str(Path(relative).parent).replace('\\', '/')
# 1) Read current contents via manage_script.read to compute SHA and true EOF location
contents = None
read_success = False
for attempt in range(3):
writer = None
try:
reader, writer = await asyncio.wait_for(asyncio.open_connection(host, port), timeout=TIMEOUT)
await asyncio.wait_for(do_handshake(reader), timeout=TIMEOUT)
read_payload = {
"type": "manage_script",
"params": {
"action": "read",
"name": name_base,
"path": dir_path
}
}
await write_frame(writer, json.dumps(read_payload).encode("utf-8"))
resp = await asyncio.wait_for(read_frame(reader), timeout=TIMEOUT)
read_obj = json.loads(resp.decode("utf-8", errors="ignore"))
result = read_obj.get("result", read_obj) if isinstance(read_obj, dict) else {}
if result.get("success"):
data_obj = result.get("data", {})
contents = data_obj.get("contents") or ""
read_success = True
break
except Exception:
# retry with backoff
await asyncio.sleep(0.2 * (2 ** attempt) + random.uniform(0.0, 0.1))
finally:
if 'writer' in locals() and writer is not None:
try:
writer.close()
await writer.wait_closed()
except Exception:
pass
if not read_success or contents is None:
stats["apply_errors"] = stats.get("apply_errors", 0) + 1
await asyncio.sleep(0.5)
continue
# Compute SHA and EOF insertion point
import hashlib
sha = hashlib.sha256(contents.encode("utf-8")).hexdigest()
lines = contents.splitlines(keepends=True)
# Insert at true EOF (safe against header guards)
end_line = len(lines) + 1 # 1-based exclusive end
end_col = 1
# Build a unique marker append; ensure it begins with a newline if needed
marker = f"// MCP_STRESS seq={seq} time={int(time.time())}"
seq += 1
insert_text = ("\n" if not contents.endswith("\n") else "") + marker + "\n"
# 2) Apply text edits with immediate refresh and precondition
apply_payload = {
"type": "manage_script",
"params": {
"action": "apply_text_edits",
"name": name_base,
"path": dir_path,
"edits": [
{
"startLine": end_line,
"startCol": end_col,
"endLine": end_line,
"endCol": end_col,
"newText": insert_text
}
],
"precondition_sha256": sha,
"options": {"refresh": "immediate", "validate": "standard"}
}
}
apply_success = False
for attempt in range(3):
writer = None
try:
reader, writer = await asyncio.wait_for(asyncio.open_connection(host, port), timeout=TIMEOUT)
await asyncio.wait_for(do_handshake(reader), timeout=TIMEOUT)
await write_frame(writer, json.dumps(apply_payload).encode("utf-8"))
resp = await asyncio.wait_for(read_frame(reader), timeout=TIMEOUT)
try:
data = json.loads(resp.decode("utf-8", errors="ignore"))
result = data.get("result", data) if isinstance(data, dict) else {}
ok = bool(result.get("success", False))
if ok:
stats["applies"] = stats.get("applies", 0) + 1
apply_success = True
break
except Exception:
# fall through to retry
pass
except Exception:
# retry with backoff
await asyncio.sleep(0.2 * (2 ** attempt) + random.uniform(0.0, 0.1))
finally:
if 'writer' in locals() and writer is not None:
try:
writer.close()
await writer.wait_closed()
except Exception:
pass
if not apply_success:
stats["apply_errors"] = stats.get("apply_errors", 0) + 1
except Exception:
pass
await asyncio.sleep(1.0)
async def main():
ap = argparse.ArgumentParser(description="Stress test the Unity MCP bridge with concurrent clients and reload churn")
ap.add_argument("--host", default="127.0.0.1")
ap.add_argument("--project", default=str(Path(__file__).resolve().parents[1] / "TestProjects" / "UnityMCPTests"))
ap.add_argument("--unity-file", default=str(Path(__file__).resolve().parents[1] / "TestProjects" / "UnityMCPTests" / "Assets" / "Scripts" / "LongUnityScriptClaudeTest.cs"))
ap.add_argument("--clients", type=int, default=10)
ap.add_argument("--duration", type=int, default=60)
ap.add_argument("--storm-count", type=int, default=1, help="Number of scripts to touch each cycle")
args = ap.parse_args()
port = discover_port(args.project)
stop_time = time.time() + max(10, args.duration)
stats = {"pings": 0, "menus": 0, "mods": 0, "disconnects": 0, "errors": 0}
tasks = []
# Spawn clients
for i in range(max(1, args.clients)):
tasks.append(asyncio.create_task(client_loop(i, args.host, port, stop_time, stats)))
# Spawn reload churn task
tasks.append(asyncio.create_task(reload_churn_task(args.project, stop_time, args.unity_file, args.host, port, stats, storm_count=args.storm_count)))
await asyncio.gather(*tasks, return_exceptions=True)
print(json.dumps({"port": port, "stats": stats}, indent=2))
if __name__ == "__main__":
try:
asyncio.run(main())
except KeyboardInterrupt:
pass