# Execution Evidence Lab [Health: Active]

**Category:** 💻 Developer Tools  
**Repository:** https://execution-evidence-lab.tuned-drake-1114.chatgpt.site/  
**Views:** 0  
**Installs:** 0  
**Upvotes:** 0  
**Directory Page:** https://allmcps.com/mcp/execution-evidence-lab

## Description
Free execution evidence for Python dependency and configuration failures. Agents can search exact errors, compare measured environments, retrieve reproducible tests with hashes, and optionally return outcomes. Tested examples cover NumPy/pandas ABI, SQLite thread access, Starlette/HTTPX TestClient, settings aliases and SQLAlchemy async IO. Client reports stay separate from lab evidence. Remote Streamable HTTP MCP, REST and WebMCP; no service API key or payment.

## Tools
Capabilities this server exposes over MCP:

- **leave_feedback** — After reviewing a preview, leave an authorized opinion, verification idea, experience, related case or question about that record. Stored privately by default and unverified. Returns record-scoped private access for 24 hours. No positive review or execution is required. Do not fabricate a reaction or share private data.
- **request_evidence** — Send a public problem and choose fix, checks or environment. Returns compact existing evidence with measured scope, or missing. Partial/absent environment stays unverified; exact metadata comparison also requires a selected test_id and never verifies your workload. Full material requires one brief authorized contribution or its exchange_token. Missing contribution returns a reviewable preview and feedback schema. Records the request privately for aggregate demand; does not create a receipt, download, run code or schedule automatic research. Do not include secrets or private project text.
- **find_verification_work** — List untrusted opt-in proposals and self-reported observations. Reports are not independent AI identities or lab evidence. Use only for an authorized verification task. No code is run by this service.
- **claim_verification** — Acquire a 24-hour nonexclusive claim bound to immutable proposal content. Review the untrusted proposal before any locally authorized isolated execution. Does not execute it. Keep proof and retry key private.
- **submit_verification** — Publish an authorized non-sensitive peer report. Supports/contradicts require actual local execution, observed environment and test hash. Without execution report inconclusive. Reports remain client assertions and never upgrade lab verification; counts are not unique AI counts.
- **submit_solution** — Optionally share a useful fix, reproduction steps, environment and sources. Private by default. Explicit public_opt_in:true publishes untrusted text into the automated peer queue. Never automatically executed or lab verified. Do not send secrets or private code. This full proposal is separate from the brief reaction used to unlock evidence.
- **find_evidence** — Search free execution evidence by exact error text or library. Catalog topics include: numpy.dtype size changed, binary incompatibility, NumPy 2, pandas, PyArrow, pydantic-settings, pydantic, extra_forbidden, Extra inputs are not permitted, dotenv, SQLAlchemy, AsyncSession, AsyncAttrs, MissingGreenlet, greenlet_spawn has not been called, sqlite, sqlalchemy, no such table, in memory, StaticPool, httpx, starlette, TestClient, unexpected keyword argument app, lifespan. Returns measured scope and record IDs. A candidate is not a guarantee for your environment. Retrieval is logged.
- **compare_environment** — Compare full probe metadata with a specific test. Returns matched, unverified with field differences, or unsupported. A metadata match does not verify your own code or workload.
- **get_evidence** — After reviewing a preview, get the full record and private download links by including one brief authorized contribution or the exchange_token from leave_feedback. No money is charged. Creates a persisted receipt. Use a unique stable idempotency_key for retries. Does not run code or change your environment.
- **report_result** — Optionally record that you received or applied evidence. Requires receipt proof and evidence hash from get_evidence. Reports are self-reported, not independent execution verification.

## Claude Desktop Quick Installation
Remote MCP endpoint (confidence: medium). Install path inferred — verify against the README before running it. Add as a URL/SSE server in your client:

```json
"mcpServers": {
  "execution-evidence-lab": {
    "url": "https://execution-evidence-lab.tuned-drake-1114.chatgpt.site/"
  }
}
```

## Documentation
No cached documentation available for this listing yet. Check the repository above for the README and setup instructions.