n8n-MCP

// Overview

An AI assistant asked to build an n8n workflow has to guess at node names, parameter shapes and defaults, and it guesses badly. n8n-MCP closes that gap: it is a Model Context Protocol server that hands the assistant structured, queryable knowledge about n8n itself — 2,755 nodes (832 core plus 1,923 community, 1,591 of them verified), node properties at 99% schema coverage, operations at 66.5%, and documentation at 86% coverage pulled from the official n8n docs.

It also carries the material that makes a generated workflow plausible rather than merely syntactically valid: 267 AI-capable tool variants with full documentation, 156 ranked real-world configurations extracted from popular templates, and a library of 2,352 workflow templates with 99.96% AI-metadata coverage. Templates can be searched by metadata, by task, by node type or by plain keyword, which is why the project's recommended system prompt tells the assistant to check templates before building anything from scratch, and to run a minimal-then-full node validation followed by a whole-workflow validation.

There are two ways to run it. The hosted dashboard needs no installation and has a free tier of 100 tool calls a day; self-hosting is a plain npx, Docker, Railway or local install. Adding an n8n API URL and key unlocks the management tools that talk to your own instance, and there is separate support for instance-level MCP, Cloudflare Access service tokens and deployment inside n8n itself via the MCP Client Tool node. Setup guides ship for Claude Code, VS Code with Copilot, Cursor, Windsurf, Codex and Antigravity.

The project is MIT-licensed TypeScript, tracks n8n releases closely (the README badge pins the n8n version it is built against), and carries a blunt safety warning worth repeating: never point it at production workflows directly. Copy the workflow, test in development, export backups and validate before deploying.

// What it does

• 2,755 n8n nodes indexed — 832 core and 1,923 community (1,591 verified) — with 99% property-schema coverage and 86% documentation coverage
• 2,352 workflow templates searchable by metadata, task, node type or keyword, plus 156 ranked real-world node configurations
• 267 AI-capable tool variants detected and documented, so assistants can wire n8n's AI nodes correctly
• Multi-level validation: validate_node (minimal, then full) followed by validate_workflow before anything is deployed
• Optional management tools against your own n8n instance via N8N_API_URL / N8N_API_KEY, including Cloudflare Access service-token auth
• Hosted service with a 100-call/day free tier, or self-host via npx, Docker, Railway or a local install
• Documented setup for Claude Code, VS Code + Copilot, Cursor, Windsurf, Codex and Antigravity

// Getting started

The fastest route is the hosted dashboard at dashboard.n8n-mcp.com — sign up, take the API key, point your MCP client at it. To self-host, you only need Node.js installed; the npm package ships a pre-built database of all the node information.

npx n8n-mcp

{
  "mcpServers": {
    "n8n-mcp": {
      "command": "npx",
      "args": ["n8n-mcp"],
      "env": {
        "MCP_MODE": "stdio",
        "LOG_LEVEL": "error",
        "DISABLE_CONSOLE_OUTPUT": "true"
      }
    }
  }
}

{
  "mcpServers": {
    "n8n-mcp": {
      "command": "npx",
      "args": ["n8n-mcp"],
      "env": {
        "MCP_MODE": "stdio",
        "LOG_LEVEL": "error",
        "DISABLE_CONSOLE_OUTPUT": "true",
        "N8N_API_URL": "https://your-n8n-instance.com",
        "N8N_API_KEY": "your-api-key"
      }
    }
  }
}

tools_documentation()
search_templates({searchMode: 'by_task', task: 'webhook_processing'})
validate_node(mode='minimal') -> validate_node(mode='full') -> validate_workflow

Commands and code are distilled from the project's own documentation — always check the official repo for the latest.

// When to use it

• Letting Claude Code, Cursor or Windsurf assemble an n8n workflow from a description without inventing node names or parameters
• Finding an existing template for a task instead of generating a workflow from zero
• Validating a generated workflow node-by-node before it ever reaches an n8n instance
• Giving an agent read and management access to your own n8n deployment, behind Cloudflare Access if you use it

// How n8n-MCP compares

n8n-MCP alongside other open-source workflow automation tools AI/TLDR tracks, ranked by GitHub stars.