claude-faf

This MCP server implements the Claude FAF (FAF) context management. It exposes a 33-tool MCP suite designed to initialize, synchronize, and optimize AI context for a codebase using the .faf/.CLAUDE.md workflow. Core capabilities include generating and reading project DNA, bi-sync between project.faf and CLAUDE.md, and a suite of utilities to assess AI-readiness and maintain context across sessions. Users typically run the server via npx claude-faf-mcp in their MCP configuration, which enables Claude-based tools to operate within the established project DNA and keep AI context aligned with the repository.

Once the MCP server is running, you can invoke tools such as faf_init to initialize your project DNA, faf_score to measure AI-readiness (0-100%), faf_sync to bi-sync between .faf and CLAUDE.md, and faf_quick to rapidly scaffold new context or files. The full toolkit is designed to work standalone without extra CLI dependencies, delivering fast, reliable context management for AI assistants across development workflows. The server supports both local and team-wide usage via MCP clients, enabling consistent AI behavior across developers and environments.

Prerequisites:

• Node.js and npm installed on your system
• Basic familiarity with MCP configuration and running npm-based MCP servers

Step-by-step installation:

1. Install Node.js from https://nodejs.org/ (includes npm)
2. Install the MCP server globally using npm:

Code:

npm install -g claude-faf-mcp

3. Verify installation:

claude-faf-mcp --version

4. Update your MCP configuration to point to the installed server. Example (in your MCP config):

{
  "mcpServers": {
    "claude-faf": {
      "command": "npx",
      "args": ["-y", "claude-faf-mcp"]
    }
  }
}

5. Start using the server via your MCP client by referencing the server name (claude-faf). You can also trigger via CLI examples provided by the package when needed.

Tips and considerations:

• The server uses the npm package claude-faf-mcp; ensure you have network access to install and run it.
• The recommended command to run in MCP configs is: {"command": "npx", "args": ["-y", "claude-faf-mcp"]} and then reference the server as claude-faf in your mcpServers mapping.
• If you need to customize behavior, you can add environment variables under env in the mcpServers configuration (e.g., for debugging or logging).
• The 33 MCP tools operate standalone with no extra CLI dependencies beyond the MCP host, and they are designed for speed and reliability in maintaining AI-context DNA.
• Ensure your project contains the canonical FAF files (project.faf and CLAUDE.md) to leverage full bi-sync capabilities.
• For CI environments, consider pinning the exact version of claude-faf-mcp to avoid breaking changes.