TL;DR: gemini-faf-mcp v2.1.1 ships 12 MCP tools powered by faf-python-sdk, built on FastMCP. The headline: faf_auto scans your project, detects your stack from manifest files, and generates a .faf — zero to project DNA in one tool call. Install from the Gemini Extensions Gallery or PyPI.

faf_auto: Zero to Project DNA

The new faf_auto tool scans your project directory, reads manifest files, and detects your stack automatically. No manual input. No guessing.

Detects From

pyproject.toml — Python, build system, deps
package.json — JavaScript/TypeScript, framework
Cargo.toml — Rust, cargo
go.mod — Go, go modules
requirements.txt — Python fallback
Gemfile — Ruby, bundler
composer.json — PHP, composer

Outputs

Language, framework, build tool
Package manager, database, API type
Score and tier
New .faf or updated existing

In Gemini CLI, just say:

> Auto-detect my project and create a .faf file

Gemini calls faf_auto. Your project gets a scored, validated .faf. Done.

The Tools

Tool	What It Does
`faf_auto`	Auto-detect stack and generate/update .faf
`faf_init`	Create a starter .faf file
`faf_read`	Parse a .faf file into structured data
`faf_validate`	Validate — score, tier, errors, warnings
`faf_score`	Quick score check (0-100%) with tier
`faf_discover`	Find .faf files in the project tree
`faf_stringify`	Convert FAF data back to YAML
`faf_context`	Gemini-optimized context from .faf
`faf_gemini`	Export GEMINI.md
`faf_agents`	Export AGENTS.md
`faf_model`	100% Trophy-scored example .faf for 15 project types
`faf_about`	FAF format info, IANA registration

Every tool delegates to faf-python-sdk for parsing, validation, and discovery. The server is pure Python — no shelling out, no Node dependencies.

Using with Gemini CLI

No commands to memorize. Just talk to it:

> Auto-detect my project and create a .faf file
> Read my project DNA
> What's the FAF score for this project?
> Export a GEMINI.md
> Show me a 100% example for a web app
> Create an AGENTS.md for this project

Gemini picks the right tool. You get the result.

Architecture

gemini-faf-mcp v2.1.1
├── server.py              → FastMCP MCP server (12 tools)
├── models.py              → 15 Trophy-scored .faf examples
├── main.py                → Cloud Run REST API (GET/POST/PUT)
└── src/gemini_faf_mcp/    → Python SDK (FAFClient, parser)

The MCP server handles local .faf operations. The Cloud Run API handles live badges, multi-agent context brokering, and voice-to-FAF mutations. Both ship in the same package.

Built on FastMCP. Powered by faf-python-sdk.

Install

gemini extensions install https://github.com/Wolfe-Jam/gemini-faf-mcp

Or from PyPI:

pip install gemini-faf-mcp

PyPI

Install the latest version.

gemini-faf-mcp

GitHub

Source, tests, and release notes.

v2.1.1 Release

Testing

183 tests across two suites:

126 MCP server tests — WJTTC 9-tier championship suite (Brake, Engine, Aero, Scoring, Exports, Safety, Contract, Roundtrip, Gallery)
57 Cloud Function tests — 7 tiers + integration

pip install -e ".[dev]"
python -m pytest tests/ -v

The Numbers

v2.1.1 — Released March 8, 2026
183/183 — Tests passing
12 — MCP tools
7 — Manifest file types detected
15 — Trophy-scored example .faf models
Python 3.10+ — Works anywhere pip does

The Ecosystem

Package	Platform	Registry
claude-faf-mcp	Anthropic	npm + MCP #2759
gemini-faf-mcp	Google	PyPI
grok-faf-mcp	xAI	npm
rust-faf-mcp	Rust	crates.io
faf-cli	Universal	npm