node-red-mcp

# Node-RED MCP Server A **MCP (Model Context Protocol)** server for managing remote Node-RED services, supporting operations such as adding, deleting, modifying, and querying Node-RED flows through natural language in MCP clients like Dify, Claude Desktop, and Cursor. ## ✨ Features | Tool | Description | | --- | --- | | `list_flows` | List all flow (tab) summaries | | `get_flow` | Get the complete configuration of a single flow | | `get_all_flows_raw` | Get all flows' raw configurations (including rev) | | `create_flow` | Create a new flow | | `update_flow` | Update a flow (name, description, nodes, enable/disable) | | `delete_flow` | Delete a flow | | `enable_flow` / `disable_flow` | Enable / disable a flow | | `deploy_flows` | Overall deployment (equivalent to the Deploy button in the editor) | | `list_node_types` | List installed node modules and types | | `get_runtime_settings` | View Node-RED runtime settings | Built on top of the [Node-RED Admin API](https://nodered.org/docs/api/admin/methods/). ## 📦 Installation ```bash python -m venv venv venv\Scripts\activate # Windows # source venv/bin/activate # Linux / macOS pip install -r requirements.txt ``` ## ⚙️ Configuration Copy `.env.example` to `.env` and fill in the actual values: ```bash cp .env.example .env ``` ```dotenv NODE_RED_URL=http://your-node-red:1880 NODE_RED_TOKEN= # Recommended: use Bearer Token directly NODE_RED_USERNAME= # Or use username and password (when adminAuth is enabled) NODE_RED_PASSWORD= ``` ### How to obtain the Node-RED Token? If Node-RED has `adminAuth` enabled, you can obtain the token by: ```bash curl -X POST http://your-node-red:1880/auth/token \ -d "client_id=node-red-admin" \ -d "grant_type=password" \ -d "scope=*" \ -d "username=admin" \ -d "password=xxx" ``` The returned `access_token` can be used as `NODE_RED_TOKEN`. This project also supports directly configuring the username and password, and automatically exchanging tokens at startup. ## 🚀 Quick Verification First, confirm that you can connect to Node-RED: ```bash python flow_get.py ``` Normally, it should output a list of flows. ## ▶️ Running the MCP Server ### 1. stdio Mode (local, for Cursor / Claude Desktop) ```bash python server.py ``` Add the following to `~/.cursor/mcp.json` in Cursor: ```json { "mcpServers": { "node-red": { "command": "python", "args": ["C:/02Code/Node-RED-MCP-Server/server.py"], "env": { "NODE_RED_URL": "http://your-node-red:1880", "NODE_RED_TOKEN": "xxx" } } } } ``` ### 2. SSE Mode (for Dify) ```bash python server.py --transport sse --host 0.0.0.0 --port 8765 ``` In Dify, go to Tools → MCP → Add MCP Service: - Server URL: `http://your-host:8765/sse` - Transport Type: `SSE` ### 3. Streamable HTTP Mode ```bash python server.py --transport http --host 0.0.0.0 --port 8765 ``` Enter the URL: `http://your-host:8765/mcp`. ## 🧩 Usage Examples (calling through large language models with natural language) - "List all flows on Node-RED" - "Disable the flow named 'Temperature Alarm'" - "Create a new flow named 'MQTT Collection' with the description 'xxx'" - "Rename the flow with id `abc123` to 'Data Cleaning'" - "Delete the flow with id `abc123`" ## 📁 Project Structure ``` . ├── server.py # MCP Server main entry ├── node_red_client.py # Node-RED Admin API client wrapper ├── flow_get.py # Connection test script ├── requirements.txt ├── .env.example └── README.md ``` ## ⚠️ Notes - **Overall deployment will overwrite existing configurations**, so it's recommended to backup using `get_all_flows_raw` first. - The `nodes` passed in `update_flow` will **replace** all nodes under the tab, not merge. - Node-RED v3+ uses flows API v2 by default, and the client has automatically carried the corresponding request header. - It's recommended to configure a reverse proxy + HTTPS + authentication for the MCP Server to avoid exposing the Node-RED management interface.