The Eunomia MCP Server extends the Eunomia framework to orchestrate data governance policies within MCP-powered text streams. It allows you to define an Orchestra—composed of instruments that implement governance rules (for example PII handling, access restrictions, or content moderation)—and apply those policies to requests flowing through MCP servers. This makes it possible to enforce policies across multiple connected servers and data streams in a coordinated way, so downstream components receive governance-compliant text outputs.

To use it, run the Eunomia MCP Server via uv and point it at your orchestra configuration. The server relies on the Orchestra definition to load instruments (such as PiiInstrument to redact personal data, or custom instruments you add) and applies them to incoming text according to the defined rules. You can connect multiple MCP servers under a single Eunomia orchestration, enabling centralized governance across disparate processing steps. Once running, send text through the MCP pipeline as you would with other MCP servers; Eunomia will intercept, process with its instruments, and forward the governed text to the next stage in your workflow.

Prerequisites

• Python 3.9+ (recommended) and a working Python environment
• Internet access to install dependencies
• Optional: uv (Python package) for running the server (as shown in the Integrations example)

Installation steps

1. Clone the repository: git clone https://github.com/whataboutyou-ai/eunomia-mcp-server.git
2. Create and activate a virtual environment (recommended): python -m venv venv source venv/bin/activate # on Unix/macOS venv\Scripts\activate # on Windows
3. Install dependencies (adjust package management as needed): pip install -e .

   If you rely on uv directly, ensure it is installed as well:

   pip install uv
4. Prepare your Orchestra configuration (Python code as shown in the README) and place it in your server directory.
5. Run the server: uv --directory "path/to/server/" run orchestra_server
6. Verify the server starts and is reachable via your MCP infrastructure. Adjust environment variables as needed for your deployment.

Notes

• The README example uses uv to run an orchestration server named orchestra_server. If your deployment uses a different entry point name, adjust the command accordingly.
• If you already have a Python environment manager (like pipx), you can also install uv via that method and run the same command.

Tips and common issues:

• If you encounter module import errors, ensure your virtual environment is activated and dependencies are installed in the same environment.
• The MCP server may require specific environment variables depending on the instruments you enable in your Orchestra. Start with REQUEST_TIMEOUT for tools that involve external requests and adjust as needed.
• Since this MCP server integrates Eunomia instruments, you can extend governance by adding or configuring instruments (e.g., PiiInstrument, custom detectors) within your Orchestra class.
• The README notes that this server is deprecated with respect to the latest Eunomia developments; plan for migration or compatibility checks if you rely on newer Eunomia features.