n8n MCP Server
The n8n MCP Server exposes n8n's 500+ nodes and workflow API to your AI. You can say "build a workflow that summarises yesterday's new Stripe orders into Slack every morning" and the AI looks up the node parameters, assembles the JSON, and creates the workflow.
What it really saves is documentation time — every n8n node has a different parameter shape, so building anything complex by hand means constant doc lookups. This server ships the node definitions, so the AI gets an accurate parameter schema instead of guessing.
Install / Connect
# No npm package — install from source.
# See the repo's README for the exact command:
# https://github.com/czlonkowski/n8n-mcpFollow the official README; env and args vary per server.
Why use it
Good for: people already running n8n who find canvas-dragging slow, and for having the AI debug workflows that failed.
Before connecting: it can create and modify workflows, which means giving the AI write access to your automation. Point it at a test instance first rather than production.
Similar / alternative servers
@21st-dev/magic
It's like v0, but in your Cursor / Claude Code / Windsurf: search 10,000+ React/
postman-mcp-server
Postman MCP Server — connect AI agents (Claude Code, Cursor, VS Code Copilot, Ge
app.3dstreet/3dstreet
Drive an open 3DStreet scene tab from Claude Desktop or Claude Code via MCP tool
ai.tradesearcher/mcp-server
Search TradeSearcher strategies and backtests from Claude Code, Cursor, Codex, a
Frequently Asked Questions
Does it need n8n Cloud, or does self-hosted work?
Either. It talks to n8n's API, so both cloud and self-hosted instances work — you supply the instance URL and an API key.
Can the AI actually run workflows?
Yes, it can trigger executions and read execution history to diagnose failures. That's exactly why permissions should stay conservative.