django

This Django MCP Server exposes a Model Context Protocol (MCP) surface from your Django project. By integrating django-mcp-server into your Django app, you make your models and DRF-backed endpoints available as MCP tools that AI agents and MCP clients can call. The server runs inside your Django application and serves the MCP endpoint at /mcp. Tools are declared by subclassing MCP toolsets (for example via ModelQueryToolset for model-backed queries) or by annotating DRF views to publish REST APIs as MCP tools. You can inspect declared tools with the provided manage.py mcp_inspect command and test calls via the MCP Python SDK against http://<your-host>/mcp. Claude AI and Claude Desktop can connect to this endpoint over a streamable HTTP protocol once proper OAuth2 and authentication are configured. To run locally, start the stdio server provided by django-mcp-server (as shown in the Quick Start: Run the CLI command for your environment) and ensure the server is reachable at the host/port where your Django app runs. The server supports both model-based tools (via ModelQueryToolset) and generic MCPToolset methods for arbitrary logic, including DRF-based APIs exposed as MCP tools.

Prerequisites:\n- Python 3.8+ (recommended) and a running Django project\n- pip (Python package installer)\n\nInstall the MCP server package:\nbash\npip install django-mcp-server\n\nIf you prefer installing directly from GitHub:\nbash\npip install git+https://github.com/omarbenhamid/django-mcp-server.git\n\nIntegrate into your Django project:\n1) Add the app to INSTALLED_APPS in your settings:\npython\nINSTALLED_APPS = [\n # ... your apps,\n 'mcp_server',\n]\n\n2) Include the MCP URL configuration in your project URLs (mcp endpoint defaults to /mcp):\npython\nfrom django.urls import path, include\n\nurlpatterns = [\n # your urls...\n path("", include('mcp_server.urls')),\n]\n\n3) Run the MCP server inside Django using the stdio server command (example):\nbash\npython manage.py stdio_server\n\nIf you want to run via a traditional Django runserver for development, you can also rely on your existing Django start command, but the stdio_server approach is recommended to expose an MCP-friendly stream interface.\nbash\npython manage.py runserver 0.0.0.0:8000\n

Tips and notes:\n- To publish DRF APIs as MCP tools, consider using the provided DRF publishing helpers (drf_publish_create_mcp_tool, drf_publish_update_mcp_tool, etc.).\n- For production usage, configure authentication via DJANGO_MCP_AUTHENTICATION_CLASSES and follow OAuth2 guidance if integrating with Claude or other OAuth-enabled clients.\n- If Tools are not visible, run python manage.py mcp_inspect to verify tool declarations.\n- Environment variables can configure Django MCP specifics, such as authentication and permission schemes.\n- The endpoint is typically available at /mcp within your Django project; ensure your server is reachable from clients you plan to use (e.g., localhost:8000 or a deployed domain).