Assay MCP Server

AI code verification for Claude Code, Cursor, and Windsurf.

Built on the LUCID methodology, Assay catches hallucinations, bugs, and security issues in AI-generated code — before they ship. It extracts implicit claims from code, verifies each one against the implementation, and reports exactly what would have gone to production without verification.

Quick Start

1. Get a free API key

Sign up at tryassay.ai and generate a key from your dashboard.

2. Add to your editor

Claude Code (~/.claude/settings.json):

{
  "mcpServers": {
    "assay": {
      "command": "npx",
      "args": ["-y", "assay-mcp"],
      "env": {
        "ASSAY_API_KEY": "ak_live_your_key_here"
      }
    }
  }
}

Cursor (.cursor/mcp.json):

{
  "mcpServers": {
    "assay": {
      "command": "npx",
      "args": ["-y", "assay-mcp"],
      "env": {
        "ASSAY_API_KEY": "ak_live_your_key_here"
      }
    }
  }
}

Windsurf (~/.windsurf/mcp.json):

{
  "mcpServers": {
    "assay": {
      "command": "npx",
      "args": ["-y", "assay-mcp"],
      "env": {
        "ASSAY_API_KEY": "ak_live_your_key_here"
      }
    }
  }
}

3. Use it

Ask your AI assistant to verify code:

• "Verify this file with Assay"
• "Check my code for hallucinations"
• "Generate a verified function that parses CSV files"

Tools

| Tool | Description | |------|-------------| | assay_verify | Verify a code string for hallucinations, bugs, and security issues | | assay_verify_file | Verify a file by path (auto-detects language from extension) | | assay_generate | Generate verified code from a task description | | assay_health | Check API status and validate your API key |

assay_verify

Extracts claims from your code, verifies each one, and reports what it caught.

Parameters:

• code (required) — The code to verify
• language (optional) — Programming language (auto-detected if omitted)
• context (optional) — What the code should do

assay_verify_file

Reads a file from disk and verifies it. Supports TypeScript, JavaScript, Python, Go, Rust, Java, and more.

Parameters:

• path (required) — Absolute path to the file
• context (optional) — What the code should do

assay_generate

Generates code with built-in verification. Synthesizes a spec, applies domain constraints, generates code, then verifies it — so you get code that's been checked before you see it.

Parameters:

• task (required) — What the code should do
• language (optional) — Target language

assay_health

Quick check that the API is up and your key is valid. No parameters.

What Assay catches

• Correctness bugs (logic errors, off-by-one, wrong return values)
• Security issues (injection, auth bypass, data exposure)
• Performance problems (N+1 queries, memory leaks, blocking calls)
• Missing error handling (uncaught exceptions, silent failures)
• Edge cases (null inputs, empty arrays, concurrent access)
• Type safety issues (implicit any, unsafe casts, missing checks)

Pricing

• Free: 100 verifications/month
• Pro/Team: Coming soon — join the waitlist

Links

• tryassay.ai — Dashboard, API keys, usage
• GitHub — Source code
• Benchmark Report — 100% HumanEval pass@5, +65.5% SWE-bench

License

MIT