n8n-MCP — MCP Server for AI-Assisted Workflow Automation

Introduction

n8n-MCP is a Model Context Protocol server that exposes n8n's node documentation, properties, and operations to AI assistants. It allows agents like Claude to understand every n8n integration and generate correct workflow configurations without hallucinating node parameters.

What n8n-MCP Does

• Serves structured metadata for 2,174 n8n nodes via MCP tools
• Provides node property schemas, operations, and credential requirements
• Enables AI agents to build valid n8n workflow JSON from natural language
• Surfaces node-specific documentation and parameter constraints
• Supports search and filtering across the entire n8n node catalog

Architecture Overview

The server indexes n8n's node definitions at startup, extracting properties, operations, and resource descriptions into a queryable format. When an MCP client requests information about a node, the server returns structured JSON with field names, types, default values, and descriptions. This allows the AI agent to construct syntactically correct workflow definitions.

Self-Hosting & Configuration

• Requires Node.js 18+ and npm
• Clone the repository and run npm install && npm run build
• Configure as an MCP server in Claude Desktop or other MCP clients
• No n8n instance required; node data is bundled with the server
• Update node definitions by pulling the latest version from GitHub

Key Features

• Covers all 2,174 official n8n nodes with full property details
• Returns type-safe parameter schemas including enums and conditionals
• Supports credential type lookups for authentication-heavy integrations
• Enables agents to reason about node compatibility and data flow
• Fast local startup with no external API dependencies

Comparison with Similar Tools

• n8n built-in AI — works inside n8n only; n8n-MCP works with any MCP client
• Langflow — visual AI workflow builder; n8n-MCP focuses on n8n ecosystem knowledge
• Flowise — similar to Langflow; n8n-MCP is not a workflow engine but a reference server
• Dify — full AI app platform; n8n-MCP is a focused MCP tool for n8n context

FAQ

Q: Do I need a running n8n instance? A: No. The node metadata is bundled in the server. You only need n8n when executing the workflows your agent designs.

Q: Which AI clients support it? A: Any MCP-compatible client including Claude Desktop, Claude Code, and custom MCP integrations.

Q: Does it support custom n8n nodes? A: Currently it covers official nodes. Community nodes can be added by extending the node index.

Q: How often is the node data updated? A: Updates track n8n releases. Pull the latest commit to get new node definitions.

Sources

• https://github.com/czlonkowski/n8n-mcp
• https://n8n.io/integrations