CHARM — OEM Automotive Data for AI Agents (MCP)

CHARM is an MCP server that gives AI agents authoritative OEM automotive repair data — factory service-manual part numbers, torque specs, fluid capacities, and procedures for vehicles 1982–2013, across 80+ makes.

Answers come straight from OEM factory service manuals, not a model's memory — so they're verifiable and source-attributed, not hallucinated. When a part number is identified, CHARM also returns links to purchase that exact part.

Built to be called by AI agents — Claude, ChatGPT, Gemini, and beyond.

Connect (hosted — no install)

Remote MCP endpoint (Streamable HTTP):

https://mcp.chatwithcharm.com/mcp

Claude Code

claude mcp add --transport http charm https://mcp.chatwithcharm.com/mcp

Claude Desktop — Settings → Connectors → Add custom connector → paste the URL above.

Config-file clients (Cursor, etc.)

{
  "mcpServers": {
    "charm": { "type": "http", "url": "https://mcp.chatwithcharm.com/mcp" }
  }
}

Then just ask naturally:

What's the front brake caliper part number for a 2009 Ford Taurus X, and where can I buy it?

What's the spark plug gap for a 2001 Toyota Camry?

Tools

Coverage

• Model years 1982–2013 — the DIY-repair sweet spot (out-of-warranty, owner-serviced vehicles).
• 80+ makes.
• Data sourced directly from OEM factory service manuals — authoritative, not generated.

Self-hosting (optional)

You don't need to self-host to use CHARM — the hosted endpoint above is the easy path. To run your own instance against the CHARM API, grab a free key at https://chatwithcharm.com/api/v1/docs, then:

pip install mcp httpx
export CHARM_MCP_API_KEY="your_key_here"
python3 charm_mcp_http.py        # Streamable HTTP (remote)
# or
python3 charm_mcp_server.py      # stdio (local, e.g. Claude Desktop)

License

MIT © Bitmacro