django-mcp-inspector

Starts a Model Context Protocol server that exposes Django runtime information as resources, allowing AI agents to inspect a live Django project without any static file analysis.

Resources

URI	Contents
django://settings	All settings (sensitive values redacted)
django://apps	Installed apps, labels, paths, model names
django://urls	URL pattern tree flattened to pattern / view / name
django://models	Every model's fields, types, and relations
django://models/{app_label}	Fields for a single app
django://migrations	Applied and pending migrations per app

Usage

Zero project modification (recommended)

Run as an ephemeral tool — no changes to INSTALLED_APPS or requirements:

DJANGO_SETTINGS_MODULE=myproject.settings uv run --with django-mcp-inspector django-mcp-inspector

Pass --settings instead of the environment variable if you prefer:

uv run --with django-mcp-inspector django-mcp-inspector --settings myproject.settings

Exclude third-party app noise:

uv run --with django-mcp-inspector django-mcp-inspector --settings myproject.settings --exclude-site-packages

As an installed app

Add mcp_inspector to INSTALLED_APPS:

INSTALLED_APPS = [
    ...
    'mcp_inspector',
]

Then run the management command:

python manage.py mcpserver
python manage.py mcpserver --exclude-site-packages

Connecting an MCP client

The server communicates over stdio. Point any MCP-compatible client (Claude Desktop, custom agent, etc.) at the process.

Example Claude Desktop config (claude_desktop_config.json):

{
  "mcpServers": {
    "django": {
      "command": "uv",
      "args": [
        "run", "--with", "django-mcp-inspector",
        "django-mcp-inspector", "--settings", "myproject.settings"
      ],
      "env": {
        "PYTHONPATH": "/path/to/your/project"
      }
    }
  }
}

Requirements

• Python 3.12+
• Django 4.2+
• mcp 1.0+