faf

faf-mcp is a universal MCP server designed to synchronize AI context across multiple platforms (Cursor, Windsurf, Cline, VS Code, Claude, and more). It hosts the core FAF tools in a single, consistent format so that any MCP client can connect and retrieve a unified context. Once running, you can use the included tools to initialize FAF project files, measure AI readiness, sync context across platforms, and perform interop operations like importing/exporting agent and cursor rules files. The server exposes endpoints and a transport for delivering real-time context updates to connected clients via MCP-SSE and supports a local development workflow with npx faf-mcp for quick testing and iteration.

Prerequisites:

• Node.js (recommended LTS) installed on your machine
• Basic terminal/command-line experience

Installation steps:

1. Install the FAF MCP package and run locally:

npx faf-mcp

2. Alternatively, install globally and run from anywhere:

npm install -g faf-mcp
faf-mcp

3. If you want to predefine the MCP server in your own config, you can create an MCP config file (e.g., mcp.json) and reference the server:

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

Then run using your preferred runtime according to your environment.

Tips and notes:

• The server is Node.js-based and uses the npx workflow for on-demand execution, which keeps your environment lean.
• Use the Local deployment example to quickly spin up and test in your development environment: npx faf-mcp
• If you customize mcp config, ensure the tool names (faf, faf-mcp) align with your client configuration and file paths.
• The project supports multiple platforms; ensure your MCP clients point to the correct transport endpoints (e.g., /sse for Server-Sent Events) as described in the README.
• When troubleshooting, verify Node.js version compatibility and network accessibility between the MCP server and clients.