mcp-builder

Use the MCP Python SDK with FastMCP to register tools via decorators, define Pydantic v2 models for input validation, and use async/await for all I/O. The Skill's Python implementation guide covers server initialization, tool registration, and a quality checklist.

MCP servers are long-running processes, so running them directly blocks your process. Use the provided evaluation harness which manages the server lifecycle for stdio transport, run the server in tmux, or use a timeout like 'timeout 5s python server.py'.

Stdio runs the server as a local subprocess for single-client CLI use, HTTP supports multiple remote clients with request-response patterns, and SSE enables server-to-client streaming for real-time updates. Choose based on deployment needs and client count.

Create an XML file with 10 complex, read-only questions having single verifiable answers, then run scripts/evaluation.py against your server. The harness reports accuracy, tool call counts, and agent feedback on tool design.

Design tools around complete workflows rather than raw API endpoints, return concise high-signal data respecting context limits, and write error messages that suggest corrective actions. Use service-prefixed snake_case names like github_create_issue and add annotations such as readOnlyHint.