Django Firebase
STREAMABLE HTTPDjango app implementing Firebase MCP server for AI agents to interact with Firebase services.
Django app implementing Firebase MCP server for AI agents to interact with Firebase services.
A comprehensive Django app that implements Firebase Model Context Protocol (MCP) server, enabling AI agents to interact with Firebase services through a standardized protocol.
Get up and running in under 5 minutes with the standalone Firebase agent for testing.
git clone https://github.com/your-repo/django-firebase-mcp.git cd django-firebase-mcp
pip install -r requirements.txt
credentials.json in the project rootMake sure these services are enabled in your Firebase project:
Create a .env file in the project root:
# Firebase Configuration SERVICE_ACCOUNT_KEY_PATH=credentials.json FIREBASE_STORAGE_BUCKET=your-project-id.appspot.com # MCP Configuration MCP_TRANSPORT=http MCP_HOST=127.0.0.1 MCP_PORT=8001 # Django Settings DEBUG=True SECRET_KEY=your-secret-key-here
⚠️ Important: Replace your-project-id with your actual Firebase project ID.
The Firebase MCP agent uses state management for conversation persistence. Choose one option:
Install and run Redis on port 6379:
# Install Redis using Chocolatey choco install redis-64 # Or download from: https://github.com/microsoftarchive/redis/releases # Then run Redis server redis-server
Add to your .env file:
# Redis Configuration REDIS_URL=redis://localhost:6379 USE_REDIS=true
For quick testing without Redis, the agent will automatically use InMemorySaver. No additional setup required.
Add to your .env file:
# Memory-based state (no persistence) USE_REDIS=false
Note: InMemorySaver doesn't persist conversations between restarts, while Redis maintains state across sessions.
Test your setup immediately with the standalone Firebase agent:
# Run the standalone agent python firebase_admin_mcp/standalone_firebase_agent.py
You should see:
🔥 Firebase MCP Agent Ready!
Type 'help' for available commands, 'quit' to exit.
>
Try these commands:
> List all Firebase collections
> Check Firebase health status
> help
> quit
For full Django integration:
# Apply migrations python manage.py migrate # Create superuser (optional) python manage.py createsuperuser # Run Django development server python manage.py runserver 8001
The MCP server will be available at: http://127.0.0.1:8001/mcp/
# Run standalone Firebase agent (quick testing) python firebase_admin_mcp/standalone_firebase_agent.py # Run MCP server via Django python manage.py runserver 8001 # Run MCP server in stdio mode (for MCP clients) python manage.py run_mcp --transport stdio # Run MCP server in HTTP mode python manage.py run_mcp --transport http --host 127.0.0.1 --port 8001 # Run standalone agent via Django management command python manage.py run_standalone_agent
# Test Firebase connectivity python firebase_admin_mcp/tests/test_firebase_connection.py # Test MCP server completeness python firebase_admin_mcp/tests/test_mcp_complete.py # Demo Firebase agent python firebase_admin_mcp/tests/demo_firebase_agent.py # Demo standalone agent python firebase_admin_mcp/demo_standalone_agent.py
The MCP server provides 14 Firebase tools across three categories:
firebase_verify_token - Verify Firebase ID tokensfirebase_create_custom_token - Create custom auth tokensfirebase_get_user - Get user info by UIDfirebase_delete_user - Delete user accountsfirestore_list_collections - List all collectionsfirestore_create_document - Create new documentsfirestore_get_document - Retrieve documentsfirestore_update_document - Update documentsfirestore_delete_document - Delete documentsfirestore_query_collection - Query with filtersstorage_list_files - List files with filteringstorage_upload_file - Upload filesstorage_download_file - Download filesstorage_delete_file - Delete filescurl http://127.0.0.1:8001/mcp/
curl -X POST http://127.0.0.1:8001/mcp/ \ -H "Content-Type: application/json" \ -d '{ "jsonrpc": "2.0", "method": "tools/call", "params": { "name": "firestore_list_collections", "arguments": {} }, "id": 1 }'
from langchain_openai import ChatOpenAI from langgraph.prebuilt import create_react_agent # Import Firebase tools from firebase_admin_mcp.tools.agents.firebase_mcp_client import ALL_FIREBASE_TOOLS # Create agent with Firebase capabilities model = ChatOpenAI(model="gpt-4") agent = create_react_agent( model=model, tools=ALL_FIREBASE_TOOLS, prompt="You are a Firebase assistant with full database and storage access." ) # Use the agent response = agent.invoke({ "messages": [{"role": "user", "content": "Show me all my Firestore collections"}] })
This project includes comprehensive documentation:
FIREBASE_ADMIN_MCP.md - Complete technical documentation
STANDALONE_AGENT.md - Standalone agent documentation
Problem: Default app does not exist error
Solution: Verify credentials.json path in .env file
Problem: Server won't start
Solution: Check if port 8001 is available: netstat -an | findstr :8001
Problem: Firebase connection fails Solution: Verify Firebase services are enabled in console
Problem: Import errors
Solution: Ensure all dependencies installed: pip install -r requirements.txt
Problem: Redis connection fails
Solution: Verify Redis is running: redis-cli ping (should return "PONG")
Problem: State not persisting between sessions Solution: Check Redis configuration or switch to Redis from InMemorySaver
django-firebase-mcp/
├── README.md # This file
├── FIREBASE_ADMIN_MCP.md # Complete documentation
├── STANDALONE_AGENT.md # Standalone agent guide
├── requirements.txt # Python dependencies
├── credentials.json # Firebase credentials (you create this)
├── .env # Environment variables (you create this)
├── manage.py # Django management
├── firebase_admin_mcp/ # Main MCP app
│ ├── standalone_firebase_agent.py # Standalone agent
│ ├── tools/ # Firebase MCP tools
│ ├── management/commands/ # Django commands
│ └── tests/ # Test suite
└── django_firebase_mcp/ # Django project settings
MIT License - see LICENSE file for details.
🔥 Ready to supercharge your AI agents with Firebase?
Start with the standalone agent, then explore the full documentation for advanced usage!