A contractor verification MCP server for AI agents.

Connect

Hosted endpoint (streamable HTTP).

1. Create a key: POST https://api.bettervetters.com/v1/keys with body {"label":"mcp"}. It includes 20 free checks and is shown once.
2. Add the server to your MCP client with the URL and a Bearer header:

{
  "mcpServers": {
    "better-vetters": {
      "url": "https://api.bettervetters.com/mcp",
      "headers": {
        "Authorization": "Bearer bv_live_YOUR_KEY"
      }
    }
  }
}

Replace bv_live_YOUR_KEY with your key. Config keys differ slightly between clients; the values are the URL and the Authorization header.

Local alternative (stdio)

If you prefer to run a process yourself, download server.py, install mcp and httpx, and add:

{
  "mcpServers": {
    "better-vetters": {
      "command": "python",
      "args": [
        "/path/to/server.py"
      ],
      "env": {
        "CONTRACTOR_CHECK_API_URL": "https://api.bettervetters.com",
        "CONTRACTOR_CHECK_API_KEY": "bv_live_YOUR_KEY"
      }
    }
  }
}

Full notes: MCP README. Setup steps also appear in the docs.

Prompts

Things to ask your agent.

• “Check whether Seattle Roofing Co in Washington is licensed and pulling permits.”
• “What jurisdictions does Better Vetters cover, and how fresh is the data?”
• “Use get_coverage first, then check_contractor only if the state is covered.”
• “Look up Acme Electric in Austin, TX. Tell me the license status and the trust score, and flag it if match confidence is below 0.8.”

Those are example prompts, not results. Your agent should treat a match.confidence below 0.8 as a possible match and treat not_found as no record in the covered sources, which is not proof of a problem.