The FINOS AI Governance MCP Server (aigf) exposes a suite of MCP tools that surface and interact with AI governance framework documents stored in FINOS repositories. It discovers available frameworks, risks, and mitigations at runtime and presents them through the MCP tooling surface, enabling you to list, read, and search content via commands like list_frameworks, get_framework, search_frameworks, list_risks, get_risk, and their related mitigation tools. You can also monitor the service and its cache via get_service_health and get_cache_stats. To use the server, ensure the FINOS MCP stack is installed in a Python environment, start the MCP server (via the finos_mcp CLI entry point), and then issue tool commands against the running endpoint. The server is designed to be deployed via Docker or directly on a Python runtime, and it supports environment-based configuration for authentication and transport layers.

Prerequisites:

• Python 3.10+ installed on your host or in your container environment
• Git available to clone or fetch dependencies
• Network access to FINOS repositories for runtime discovery of frameworks, risks, and mitigations

Installation steps (local development):

1. Clone the repository git clone https://github.com/finos/aigf-mcp-server.git cd aigf-mcp-server

2. Create and activate a virtual environment (recommended) python -m venv venv

   On Windows:

   venv\Scripts\activate

   On macOS/Linux:

   source venv/bin/activate

3. Install the package in editable mode pip install -e .

4. Generate or set required environment variables (example values below; replace with secure values locally or in your deployment) export FINOS_MCP_CACHE_SECRET=$(python -c "import secrets; print(secrets.token_hex(32))") export FINOS_MCP_MCP_TRANSPORT=http export FINOS_MCP_MCP_HOST=0.0.0.0 export FINOS_MCP_MCP_PORT=8000

   Optional: authentication related vars

   export FINOS_MCP_MCP_AUTH_ENABLED=false

5. Start the MCP server (in this repository's style) finos-mcp --help mcp list tools

6. (Optional) Run via Docker or docker-compose as described in the README for containerized deployment.

Tips and common issues:

• Always set FINOS_MCP_CACHE_SECRET to a 32+ character value when enabling cache-based authentication.
• If deploying to cloud environments, pass the required FINOS_MCP_MCP_HOST, FINOS_MCP_MCP_PORT, and transport settings through your deployment platform's configuration (Secrets/Env Vars).
• When FINOS_MCP_MCP_AUTH_ENABLED=true, ensure you provide either FINOS_MCP_MCP_AUTH_JWKS_URI or FINOS_MCP_MCP_AUTH_PUBLIC_KEY along with issuer and audience values.
• For local development, using a Python virtual environment helps isolate dependencies from the system Python.
• The server discovers frameworks, risks, and mitigations at runtime; ensure the upstream repositories are accessible from your deployment environment.
• If you upgrade dependencies, re-run tests with finos-mcp --help and mcp list tools to verify tool availability after changes.