This is a Model Context Protocol (MCP) server that connects to a Coreflux MQTT broker and makes Coreflux and MQTT actions available as tools for Claude and other MCP-compatible AI assistants.

• Connects to Coreflux MQTT broker
• Provides tools for all Coreflux commands (models, actions, rules, routes)
• Discovers and lists available actions
• Includes LOT language documentation as resources
• Built with the official MCP SDK for seamless Claude integration
• Standalone setup assistant for configuration

The server includes a standalone setup assistant that can be run separately from the main server. Run the setup assistant when:

• You need to create an initial configuration (.env file)
• You want to update your existing configuration
• You're experiencing connection issues and need to reconfigure

To run the setup assistant:

python setup_assistant.py

The setup assistant helps you:

• Create or update the .env file with your configuration
• Configure MQTT broker settings (host, port, credentials)
• Set up TLS configuration if needed
• Configure logging options

After configuration is complete, you can run the server normally.

1. Create or edit ~/Library/Application Support/Claude/claude_desktop_config.json (macOS/Linux) or %USERPROFILE%\AppData\Roaming\Claude\claude_desktop_config.json (Windows)

2. Add the following configuration (adjust the paths accordingly):

   {
     "mcpServers": {
       "coreflux": {
         "command": "python",
         "args": [
           "/PATH/TO/server.py",
           "--mqtt-host", "localhost", 
           "--mqtt-port", "1883",
           "--mqtt-user", "root",
           "--mqtt-password", "coreflux",
           "--mqtt-client-id", "claude-coreflux-client",
           "--do-agent-api-key", "your_coreflux_copilot_api_key_here"
         ],
         "description": "Coreflux MQTT Broker Control",
         "icon": "🔄",
         "env": {}
       }
     }
   }

   Note: Instead of passing the API key as a command-line argument, you can set it in the .env file for better security.

   Tip: A sample claude_desktop_config.json file is included in this repository that you can use as a starting point.

3. Restart Claude Desktop

The server accepts the following command-line arguments. These settings can also be configured via the .env file using the setup assistant:

The server includes a request_lot_code tool that uses the Coreflux Copilot API to generate LOT (Logic Object Tree) code based on natural language prompts. To use this feature, you need to:

1. Get an API key from the Coreflux Copilot dashboard
2. Set the API key in your .env file:

   DO_AGENT_API_KEY=your_coreflux_copilot_api_key_here

3. Or pass it as a command-line argument:

   python server.py --do-agent-api-key your_api_key_here

Note: Without the API key, the LOT code generation feature will not work. The setup assistant will prompt you to configure this when you run it.

API Endpoint: The server is pre-configured to connect to the Coreflux Copilot API endpoint. The request format conforms to the chat completions API specification with proper authentication headers.

The server provides tools for common Coreflux commands:

• add_rule: Add a new permission rule
• remove_rule: Remove a permission rule
• add_route: Add a new route connection
• remove_route: Remove a route connection
• add_model: Add a new model structure
• remove_model: Remove a model structure
• add_action: Add a new action event/function
• remove_action: Remove an action event/function
• run_action: Run an action event/function
• remove_all_models: Remove all models
• remove_all_actions: Remove all actions
• remove_all_routes: Remove all routes
• list_discovered_actions: List all discovered Coreflux actions
• request_lot_code: Generate LOT code using Coreflux Copilot API based on natural language prompts

The MCP server now starts even if the MQTT broker is not available, allowing you to troubleshoot and configure connections through the MCP tools.

• The server will start successfully even if the MQTT broker is unreachable
• Use the get_connection_status tool to check connection health and get troubleshooting guidance
• Use the setup_mqtt_connection tool to configure a new broker connection without restarting
• Use the check_broker_health or reconnect_mqtt tools to test and retry connections

• get_connection_status: Get detailed connection status with troubleshooting guidance
• setup_mqtt_connection: Configure a new MQTT broker connection dynamically
• mqtt_connect: Connect to a specific MQTT broker with custom parameters
• check_broker_health: Test broker connectivity and attempt reconnection
• reconnect_mqtt: Force reconnection to the configured broker

If you encounter issues:

1. Verify your MQTT broker credentials in your Claude configuration
2. Ensure the broker is accessible
3. Run the setup assistant to verify or update your configuration:

   python setup_assistant.py

4. Check Claude Desktop logs:

   # Check Claude's logs for errors (macOS/Linux)
   tail -n 20 -f ~/Library/Logs/Claude/mcp*.log
   # Windows PowerShell
   Get-Content -Path "$env:USERPROFILE\AppData\Roaming\Claude\Logs\mcp*.log" -Tail 20 -Wait

5. Run the server with debug logging:

   # Direct execution with debug logging
   python server.py --mqtt-host localhost --mqtt-port 1883 --log-level DEBUG

• MCP Quickstart for Server Developers
• MCP Official Documentation