Skip to content

AGENTS.md

Latest commit

 

History

History
67 lines (48 loc) · 2.98 KB

AGENTS.md

File metadata and controls

67 lines (48 loc) · 2.98 KB
Raw

AGENTS

Audience: LLM-driven engineering agents

This file provides guidance for autonomous coding agents working inside the FastMCP repository.

Repository map

Path Purpose
src/fastmcp/ Library source code (Python ≥ 3.10)
└─server/ Server implementation, FastMCP, auth, networking
└─client/ High‑level client SDK + helpers
└─resources/ MCP resources and resource templates
└─prompts/ Prompt templates
└─tools/ Tool implementations
tests/ Pytest test‑suite
docs/ Mintlify‑flavoured Markdown, published to https://gofastmcp.com
examples/ Minimal runnable demos

Mandatory dev workflow

uv sync                              # install dependencies
uv run pre-commit run --all-files    # Ruff + Prettier + Pyright
uv run pytest                        # run full test suite

Tests must pass and lint/typing must be clean before committing.

Core MCP objects

There are four major MCP object types:

  • Tools (src/tools/)
  • Resources (src/resources/)
  • Resource Templates (src/resources/)
  • Prompts (src/prompts)

While these have slightly different semantics and implementations, in general changes that affect interactions with any one (like adding tags, importing, etc.) will need to be adopted, applied, and tested on all others. Be sure to look at not only the object definition but also the related Manager (e.g. ToolManager, ResourceManager, and PromptManager). Also note that while resources and resource templates are different objects, they both are handled by the ResourceManager.

Code conventions

  • Language: Python ≥ 3.10
  • Style: Enforced through pre-commit hooks
  • Type-checking: Fully typed codebase
  • Tests: Each feature should have corresponding tests

Development guidelines

  1. Set up the environment: 
    uv sync && uv run pre-commit run --all-files
  2. Run tests: uv run pytest until they pass.
  3. Iterate: if a command fails, read the output, fix the code, retry.
  4. Make the smallest set of changes that achieve the desired outcome.
  5. Always read code before modifying it blindly.
  6. Follow established patterns and maintain consistency.