Node-RED MCP server
A Node-RED MCP server: let an agent read, build, and update flows, manage nodes, trigger inject nodes, and inspect runtime state via the Admin API.
The Node-RED MCP Server is a community bridge between an AI assistant and the Node-RED low-code automation platform, exposing the Node-RED Admin API as MCP tools so an agent can build and operate flows in natural language. The agent can retrieve every flow or a single flow, update them, create and delete flows, list the tabs in a workspace, and read or set the deployed flow state — effectively letting it author automations the way a human would in the Node-RED editor, but driven from chat.
For working with individual nodes it can list the nodes in a flow, fetch detailed node info, search the palette, find nodes by type, and toggle node modules on or off. It can also trigger inject nodes remotely to fire a flow on demand, read the Node-RED settings, pull runtime diagnostics, and visualize or return formatted flows for the model to reason about. It runs locally over stdio via npx and points at your Node-RED instance with NODE_RED_URL, using NODE_RED_TOKEN when the editor requires authentication. Because it can deploy and delete flows, treat it as a write-capable tool and scope it to a non-production instance until you trust it.
Quick install
Copy-paste configs are provided for all 8 supported clients. Pick your client below.
Available tools
| Tool | Description |
|---|---|
| get-flows | Retrieves all flows from the Node-RED instance. |
| update-flows | Replaces all flows with a new configuration. |
| get-flow | Retrieves a single flow by ID. |
| update-flow | Updates a single flow. |
| create-flow | Creates a new flow. |
| delete-flow | Deletes a flow. |
| list-tabs | Lists the tabs (flow workspaces) in the editor. |
| get-flows-state | Returns the current runtime state of the flows. |
| set-flows-state | Sets the runtime state of the flows (start/stop). |
| get-flows-formatted | Returns flows in a formatted view for reasoning. |
| visualize-flows | Produces a visualization of the flows. |
| inject | Triggers an inject node to fire a flow on demand. |
| get-nodes | Lists the nodes in the configuration. |
| get-node-info | Returns detailed information about a node. |
| find-nodes-by-type | Finds nodes of a given type. |
| search-nodes | Searches nodes in the palette. |
| toggle-node-module | Enables or disables a node module. |
| get-settings | Reads the Node-RED settings. |
| get-diagnostics | Returns runtime diagnostics for the instance. |
| api-help | Returns help on the available Admin API operations. |
Required configuration
- NODE_RED_URLRequired
URL of your Node-RED instance (e.g. http://localhost:1880).
- NODE_RED_TOKENOptional
Node-RED Admin API access token, required when the editor has authentication enabled.
What you can do with it
Build an automation flow from a description
Describe the integration you want and the agent assembles the nodes, creates the flow, and deploys it to your Node-RED instance — then triggers an inject node to test it.
Inspect and debug a running flow
Ask the agent what a flow does and it reads the flow, the node details, and runtime diagnostics, explaining the wiring and flagging misconfigured or disabled nodes.
FAQ
- Is it free?
- Yes. node-red-mcp-server is open source under the MIT license and free to run. Note the upstream repository is archived, so it is community-maintained as-is; it works against the standard Node-RED Admin API.
- Does it support remote/OAuth?
- No. It runs locally over stdio via npx and talks to your Node-RED instance over its Admin API using NODE_RED_URL and an optional NODE_RED_TOKEN. There is no hosted OAuth endpoint.
- Can the agent change my deployed flows?
- Yes. It exposes create, update, delete, and deploy-state tools, so it is write-capable. Point it at a non-production Node-RED instance or keep deployments behind review until you trust the agent.