Architect AI Context for Codebases

Provides expert-level control over .faf (Foundational AI-context Format), an IANA-registered format (application/vnd.faf+yaml) that gives AI tools persistent project context across sessions, tools, and platforms. It covers expert configuration of .faf files and MCP servers for complex project setups, working toward a "championship scoring" tier system - Gold (95%+), Silver (85%+), and Bronze (70%+) AI-readiness scores - multi-AI workflows that share one context across Claude, Cursor, Gemini, and Windsurf, turning a legacy or archaeology-style codebase into AI-readable project context, standardized team context, and enterprise MCP server configuration.

It sets up the claude-faf-mcp MCP server (33 tools) via a simple npx-based config, and other available FAF MCP servers (grok-faf-mcp for xAI/Grok, rust-faf-mcp as a native-performance binary, gemini-faf-mcp for Gemini). It configures bi-directional sync so .faf stays synchronized with CLAUDE.md, .cursorrules, GEMINI.md, and AGENTS.md, and works with the Mk4 architecture's 33-slot format covering project identity and goals, technical stack detection, human context (who/what/why/where/when/how), architecture patterns, and deployment configuration.

# Install FAF CLI
npm install -g faf-cli

# Initialize your project
faf init

# Score AI-readiness
faf score --details

# Set up MCP server
faf mcp install

It also documents advanced CLI commands (faf score --championship --verbose, faf bi-sync --target all, faf validate --strict, faf enhance --model claude --focus completeness) and shows example .faf YAML configurations for a legacy Java/Spring enterprise system, a modern React/Vite dashboard, an enterprise multi-region microservices setup with compliance fields, and a legacy PHP-to-Laravel modernization case.

Use faf-expert for fine-tuned .faf configuration, championship-scoring optimization, multi-platform sync management, enterprise deployment patterns, and advanced MCP server setup. Use the companion faf-wizard skill instead for quick, beginner-friendly, one-click project setup with automated stack detection. As with any skill output, verify the resulting configuration, and stop and ask for clarification if required inputs, permissions, safety boundaries, or success criteria are missing.

Input is an existing or new codebase to be given AI context, or an existing .faf file to refine, sync, or score. Output is a configured .faf file and its MCP server config, a bi-directional sync setup keeping .faf aligned with tool-specific context files, and an AI-readiness score with a tier breakdown.

Integrates with Claude Code via native MCP support, Cursor via .cursorrules sync, Gemini CLI via GEMINI.md sync, and Windsurf via .windsurfrules, and is documented (per the source) at 52k+ downloads across the FAF ecosystem, 800+ tests across the CLI and MCP server, and 153+ validated project formats supported.

Developers and teams who want one persistent, standardized AI-context format shared across multiple AI coding tools, especially those needing enterprise-grade or high-scoring context configuration rather than a quick default setup.