1c_mcp

This MCP server provides tools and infrastructure to build and expose Model Context Protocol (MCP) instruments for a 1С:Enterprise base. It includes a ready-made 1С extension that handles MCP protocol details, plus an optional Python proxy to address transport and authentication challenges when connecting AI clients. The server is designed so AI models can request context data and run business logic through custom tools, resources and prompts you implement on the 1С side. The Python proxy should be used if you need to support clients that operate over stdio or require OAuth2-based authentication forwarded to 1С via Basic Auth. Client examples and configurations are provided in the mcp_client_settings folder to help you connect clients like Cursor or Claude Desktop to your HTTP service.

Prerequisites:

• Docker installed (for the recommended isolated deployment) or a compatible runtime for the chosen deployment method.
• Access to a 1С:Enterprise instance with the MCP extension built and ready (see the project documentation).
• If enabling the Python proxy, Python 3.x installed and access to the repository components under src/py_server.

Step-by-step:

1. Build or obtain the 1С MCP extension and the optional Python proxy components as described in the repository docs.
2. If using Docker (recommended for isolation):
   • Build or pull the image that contains the MCP server, e.g.: docker build -t 1c_mcp-image . (Alternatively use a prebuilt image from your registry.)
3. Prepare environment variables for the container, such as:
   • MCP_ONEC_URL: URL to the 1С HTTP service
   • MCP_ONEC_LOGIN / MCP_ONEC_PASSWORD: authentication details if used
   • MCP_PROXY_MODE: single-user or per-user mode (for proxy)
4. Run the container in detach mode: docker-compose or docker run with the appropriate env vars, e.g.: docker run -d -e MCP_ONEC_URL="https://1c.example.com" -e MCP_PROXY_MODE="single-user" -p 8080:8080 1c_mcp-image
5. Verify the MCP service is reachable at the configured endpoint and test client connections using the prepared mcp_client_settings.

Additional deployment options:

• If you skip Docker, you can run the Python proxy and MCP extension directly following the project’s README sections for the Python proxy configuration and 1С extension integration.

Tips and common notes:

• The Python proxy is recommended when your clients use stdio or when you need OAuth2-based authentication flowing to 1С via Basic Auth. See the Python proxy docs for details on configuration and docker deployment tips.
• When deploying on the same host as 1С, consider using host.docker.internal in network settings to reference the host from inside Docker.
• Review the mcp_client_settings to adapt client configurations for different AI clients (Cursor, Claude Desktop, etc.).
• Ensure the MCP extension is properly published into your 1С configuration and that the HTTP service is accessible for the MCP API backend.
• If you change network topologies or authentication flows, remember to refresh client configurations to maintain compatibility.