mxcp

MXCP (MXCP) is an enterprise-grade MCP framework designed for building production AI applications with strong data quality, security, and governance. It supports both SQL-based data queries and Python-based logic within a single MCP server, providing a cohesive environment for deploying AI tools with type safety, validation, audit trails, and policy enforcement. After installing the MXCP package, you can start the server using the included CLI and expose endpoints that leverage either SQL against your data models or Python for complex processing, ML models, or integrations. The framework also integrates open telemetry for observability and includes a testing and validation workflow to ensure your endpoints behave correctly before deployment. The README example demonstrates starting the server and connecting to it via a client like Claude Desktop, illustrating how to configure a multi-tool MCP server that serves both data queries and Python-driven tools.

Prerequisites:

• Python 3.11 or newer
• pip (comes with Python)

0. Optional: create a virtual environment python -m venv .venv source .venv/bin/activate # on macOS/Linux .venv\Scripts\activate # on Windows

1. Install MXCP from PyPI pip install mxcp

2. Verify installation mxcp --version

3. Run the MXCP server (as an example, using the recommended Python launcher) python -m mxcp serve --transport stdio # serves tools over standard I/O

4. (Optional) Create a project and bootstrap your tools mkdir my-ai-tools && cd my-ai-tools mxcp init --bootstrap

5. If you are integrating with a client, follow the README’s example to configure the mcpServers section in your client configuration and connect your tooling to the MXCP server.

Tips and common considerations:

• Ensure Python 3.11+ is installed and available in your PATH.
• Use a virtual environment to isolate dependencies.
• MXCP supports both SQL-based data endpoints and Python-based logic; structure your tool definitions accordingly and validate with mxcp validate, mxcp test, mxcp lint, and mxcp evals before deployment.
• For production deployments, consider using a process manager and enabling OpenTelemetry integration for tracing and metrics.
• If you encounter CLI issues, verify that the mxcp script is on your PATH after pip installation and try running python -m mxcp instead of the mxcp shell wrapper.
• Refer to the full production methodology and guides in the repository for best practices on data quality, security, and auditing.