How to use the MCP Generator
Generate your MCP Server Card and a real, working starter server, the two files agents need to discover and use your API.
An MCP Server Card, published at yourdomain.com/.well-known/mcp.json, is how Claude, Cursor, and other MCP clients discover that your product has an MCP server and see what it can do before connecting. The MCP Generator produces that card, plus a real, deployable Next.js starter server implementing the JSON-RPC handshake MCP clients expect, matching the exact scaffold byfaro.ai runs its own MCP endpoint on.
Running the generator
Fill in your business details and list the actions you want agents to be able to take, for example "check order status" or "search inventory". Each line you add becomes one MCP tool in your Server Card and a stub handler in your generated server code.
What gets generated
Two files, not one. First, the Server Card for /.well-known/mcp.json: it declares your server's name, description, live endpoint, and its available tools, so an agent evaluating whether to connect can see up front what it will be able to do. Second, a starter server (app/mcp/route.ts for Next.js) implementing the real JSON-RPC 2.0 handshake, initialize, ping, tools/list, and tools/call, with CORS handled. Each tool handler is an honest, clearly labeled placeholder you implement yourself; the generator doesn't fake data, because a card pointing at a broken or fake endpoint is worse than no card at all.
Deploying both files
Save the Server Card as public/.well-known/mcp.json and the server code as app/mcp/route.ts (Next.js App Router). Deploy both together, a card with no live endpoint behind it will score partial credit at best.
Verifying it worked
Faro's Agent Infrastructure category checks for both the Server Card and a live response from the endpoint it declares. A card backed by a real, responding server earns full credit; a card with a dead or missing endpoint is flagged explicitly rather than scored as if everything's fine. Re-run the AI Readiness Scan after deploying to confirm. Pair this with llms.txt (for understanding) and structured data (for verification) for the full agent-facing stack: llms.txt, structured data, MCP.