How to Set Up an n8n MCP Server with Claude (Step-by-Step)

Why MCP Changes Everything for n8n

Most AI automation is one-directional: you trigger a workflow, it runs, done. MCP (Model Context Protocol) flips this. Instead of you telling n8n what to do, your AI agent discovers and uses your workflows as tools.

Think about what that means: Claude can look at your available n8n workflows, understand what each one does, and decide which ones to call — with the right parameters — to accomplish a goal you describe in plain English.

This isn't theoretical. n8n ships with MCP server support built in since late 2025. Here's how to set it up.

What You'll Need

• n8n instance — self-hosted or n8n Cloud (v1.70+)
• Claude Desktop — or any MCP-compatible client
• 10 minutes

Step 1: Enable MCP Server in n8n

n8n's MCP server is built into the platform. You don't install anything extra.

1. Open your n8n instance and go to Settings → MCP Server
2. Toggle the MCP server ON
3. Copy the server URL — you'll need this for Claude
4. Generate an API key if you haven't already (Settings → API → Create API Key)

Your MCP server URL will look something like: http://localhost:5678/mcp (self-hosted) or https://your-instance.n8n.cloud/mcp (cloud).

Step 2: Create a Workflow with MCP Server Trigger

Not all workflows are automatically exposed to MCP. You choose which ones are available by using the MCP Server Trigger node.

1. Create a new workflow (or open an existing one)
2. Add the MCP Server Trigger node as your starting node
3. Give the trigger a clear name and description — this is what Claude sees when discovering tools
4. Define the input parameters your workflow expects
5. Build out the rest of your workflow (API calls, data transforms, etc.)
6. Activate the workflow

Step 3: Connect Claude Desktop

Now connect Claude to your n8n MCP server so it can discover and use your workflows.

1. Open Claude Desktop → Settings → Developer → MCP Servers
2. Click Add Server
3. Enter your n8n MCP server URL
4. Add your API key for authentication
5. Save and restart Claude Desktop

After reconnecting, you should see a 🔨 icon in your Claude chat indicating available tools. Click it to see all your exposed n8n workflows.

Step 4: Test It

The best test is a simple one. If you created a workflow that, say, searches a database, just ask Claude naturally:

Claude will identify the right n8n workflow, call it with the appropriate parameters, and return the results — all without you specifying which workflow to use.

Real-World Use Cases

Here's what people are actually building with n8n + MCP:

• AI-powered CRM: Claude queries your customer database, sends follow-up emails, and logs activities — all through n8n workflows
• Content pipeline: "Research this topic, write a draft, schedule it on social media" — three n8n workflows, one natural language request
• Data analysis: Claude pulls data from multiple sources via n8n, analyzes it, and generates reports
• Lead enrichment: Give Claude a company name, it runs your n8n enrichment pipeline and returns a full profile

Downloadable Starter Workflows

We've built ready-to-import n8n workflow JSONs you can use right now:

• AI Content Repurposer — Turn one piece of content into posts for every platform
• RSS to Social Media — Auto-post from any RSS feed with AI summaries
• Webhook to Notion — Capture any webhook data into structured Notion pages

Want more? Browse all templates or go Pro for premium workflows with MCP integration built in.

Common Issues

• "No tools found" — Make sure your workflow is activated and uses the MCP Server Trigger node (not a regular webhook)
• Authentication errors — Double-check your API key in Claude Desktop settings
• Claude ignores your tools — Improve your tool descriptions. Be specific about what the tool does and when to use it
• Timeouts — Long-running workflows may time out. Add a timeout node or optimize your workflow

What's Next

Once you have MCP working, the compound effects kick in. Every new n8n workflow you build automatically becomes a tool Claude can use. Your AI agent gets more capable without any additional configuration.

The teams seeing the biggest results are building libraries of focused, well-described workflows — not one mega-workflow that does everything.

Start small: one workflow, one use case, one test. Then expand.