AGENTS.md Generator — Claude Code MCP server: what it does
Parses a codebase with tree-sitter and writes an AGENTS.md file that tells AI agents how the project works. Handy the first time an agent opens a repo.
claude mcp add agents-md-generator -- uvx agents-md-generatorWhat AGENTS.md Generator does
The server reads source files in Python, C#, TypeScript, JavaScript and Go, then produces AGENTS.md following the agents.md open standard. Think of it as a README addressed to an agent rather than a person: stack overview, layers and data flow direction, naming and import conventions, environment variables, install and run commands, test and lint commands. Sections with nothing detected are left out of the file.
Three tools sit behind it. generate_agents_md runs the pipeline and hands your client chunked read instructions; scan_codebase returns the same analysis without any mandate to write a file, which is what you want when you only need architectural context; read_payload_chunk streams the result until has_more turns false. Nothing large crosses the MCP wire.
Rescans lean on a cache: only files whose SHA-256 changed get parsed again, and from those just the affected public symbols are kept. Private symbols and test file internals never enter the cache or the payload. Python 3.11+, Git and an MCP-capable client are required, so the server will not start without them. Running it through uvx skips a separate install step, though uv itself has to be present.
What it helps with
- generate AGENTS.md for a repo the agent has never seen
- pull architectural context without writing any file
- refresh the project description after code changes
- extract build and test commands from package.json or Makefile
- list environment variables found in sources and .env.example
Example prompts for the agent
- Generate the AGENTS.md for this project
- Scan the codebase and show me the architecture, don't create a file
How to install AGENTS.md Generator
Platform CLI
claude mcp add agents-md-generator -- uvx agents-md-generatoruvx runs the command named after the package; if the README names another one, use uvx --from agents-md-generator <command>
Platform CLI
codex mcp add agentsmd-generator -- uvx agents-md-generatoruvx runs the command named after the package; if the README names another one, use uvx --from agents-md-generator <command>
Config file
{
"mcpServers": {
"agentsmd-generator": {
"command": "uvx",
"args": [
"agents-md-generator"
]
}
}
}uvx runs the command named after the package; if the README names another one, use uvx --from agents-md-generator <command>
Config file
{
"servers": {
"agentsmd-generator": {
"type": "stdio",
"command": "uvx",
"args": [
"agents-md-generator"
]
}
}
}uvx runs the command named after the package; if the README names another one, use uvx --from agents-md-generator <command>
Platform CLI
gemini mcp add agentsmd-generator uvx agents-md-generatoruvx runs the command named after the package; if the README names another one, use uvx --from agents-md-generator <command>
Step by step, with commands for each platform: How to add an MCP server to Claude Code, Codex, Cursor, VS Code, Gemini CLI and more
Questions about AGENTS.md Generator
How do I install AGENTS.md Generator in Claude Code?
Run claude mcp add agents-md-generator -- uvx agents-md-generator. That registers the server for the current project only; add -s user to make it available everywhere.
What does AGENTS.md Generator do?
It parses your sources with tree-sitter and writes AGENTS.md covering stack, architecture, conventions and run commands. The same analysis is available without the file through scan_codebase.
What does the server need to run?
Python 3.11+, Git and an MCP-compatible client. The uvx route also needs uv installed.
