| name | explore-python |
| description | Research Python/FastAPI MCP server project code - tools, structure, patterns, FastAPI organization, asyncio, _meta, human-in-the-loop, OTel, Auth |
| context | fork |
| agent | Explore |
| allowed-tools | Read Grep Glob Bash |
Explore MCP Server (Python)
Explore this project's Python/FastAPI MCP server code.
Orientation — read these first
Before exploring code, check project state:
CLAUDE.md — conventions, architecture decisions, setup notesCHANGELOG.md — what's been implemented, recent changes, what's in progress
Both may be empty early in the project. Note that and proceed to code.
Service Context
- Python >= 3.13, FastAPI >= 0.115
- MCP protocol via FastMCP (
fastmcp); tool registration with @mcp.tool() async defs
_meta field used for human-in-the-loop elicitation / approval flows- Package manager uv (Astral); build backend hatchling, lockfile:
uv.lock
- Linter ruff (target
py313); type-check mypy >= 1.14
- External MCP servers in use: context7, GitHub MCP, Serena MCP, rtl (rust token killer)
What to look for
- Tool definitions (
@mcp.tool()) — inputs, outputs, _meta handling
- Human-in-the-loop patterns — elicitation requests, approval flows,
_meta passthrough
- FastAPI route organization — async handlers, dependency injection, middleware
- Auth patterns — API keys, OAuth, token validation
- Observability — OTel spans, logging, structured output
- Error handling — MCP error codes, HTTP status mapping
- Project structure — how modules are split, where config lives
Task
$ARGUMENTS
