An n8n MCP server gives an AI client deep knowledge of n8n's 1,800+ automation nodes — and, with your instance connected, the ability to create and manage workflows from a natural-language description. Here is the setup.
Step 1 — get your n8n API key
In your n8n instance: Settings → n8n API → Create an API key. Note the key and your instance URL (e.g. https://your-n8n.example.com). Self-hosted and cloud both work.
Step 2 — configure the client
The popular n8n-mcp server runs via npx:
{
"mcpServers": {
"n8n": {
"command": "npx",
"args": ["-y", "n8n-mcp"],
"env": {
"MCP_MODE": "stdio",
"N8N_API_URL": "https://your-n8n.example.com",
"N8N_API_KEY": "your_api_key"
}
}
}
}
It works in Claude Desktop, Claude Code, Cursor and Windsurf. The node-documentation tools work even without an instance; adding N8N_API_URL/N8N_API_KEY unlocks creating and managing real workflows. Restart the client; see MCP config file location.
Step 3 — verify
Ask: "Build an n8n workflow that posts new Airtable rows to Slack." It should assemble the nodes and (with API access) create the workflow.
Common problems
- Workflows aren't created — node docs work offline, but creating workflows needs a valid
N8N_API_URL+N8N_API_KEY. 401from n8n — wrong API key or URL; regenerate in Settings.- No tools in client — Node 18+ for
npx; see Cursor MCP not working.
Going further
n8n is itself an automation hub — pair the MCP with Slack, Airtable and GitHub servers. Browse the developer tools category or curated loadouts. Keep the API key private — see MCP security best practices.