Resources · n8n integration

n8n MCP Server Trigger nodeConfigure MCP Server Trigger in n8n.

The n8n MCP Server Trigger node turns a workflow into a server that AI clients can call. MCP, short for Model Context Protocol, lets a client list the available tools and run them on demand. Two parameters configure the node, and every tool you attach to it becomes callable.

Try the MCP Server Trigger node on n8nSee the 2 parameters

Verified Trustpilot reviews · AI, automation & growth agency

Why automate

What does the n8n MCP Server Trigger node do?

The MCP Server Trigger node is the entry point into n8n for MCP clients. It exposes a URL, and a client that knows this URL can list the tools attached to the node and call any of them. It is a trigger, so it opens the workflow, but it behaves differently from the others: instead of passing its output to the next node, it only connects to tool nodes and runs the one the client asked for.

First scenario, a desktop assistant that reads your business data. You attach a Custom n8n Workflow Tool to the node, that workflow queries an internal system, and the assistant calls it whenever a question needs a fresh answer. The Path parameter fixes the URL so the client configuration stays valid, and Authentication keeps strangers out.

Second scenario, one server per team. Sales gets a node whose Tool port carries a lookup into a Google Sheets pricing table, support gets another one that drafts replies through Gmail. Each node has its own path, its own credential, and the two servers never share anything.

Third scenario, one action per tool, kept small. A tool that posts to Slack, a tool that writes a row, a tool that computes something. Clients call them one at a time, so a workflow that does five things at once is harder to use than five narrow tools.

When to prefer something else. If the model lives inside n8n and you are the one orchestrating it, the AI Agent node is the right root node: it holds the model and the tools itself. The MCP Server Trigger is for the opposite direction, when an outside client drives. And when your agent needs to consume an external MCP server rather than publish one, that is the MCP Client Tool node.

The known limits are worth reading before you ship. The node supports Server-Sent Events, a long-lived connection built on HTTP, and streamable HTTP; it does not support standard input/output transport. In queue mode with a single webhook replica it works as expected, but with several replicas you have to route every /mcp* request to one dedicated webhook replica, otherwise connections break or drop events. Behind an nginx reverse proxy, proxy buffering has to be off on the MCP location, and gzip and chunked transfer encoding are usually turned off too. Last quirk: claude.ai custom connectors ask you to sign in to n8n even when Authentication is set to None, and it is the only client known to do that. If you want a wider look at the platform before building on it, the n8n review covers it.

Connect

What do you need to secure the MCP URL?

  1. 01

    Pick the authentication mode

    Open the node and look at Authentication. Four values are available: none leaves the URL open, bearerAuth expects a token in the request, headerAuth expects a named header, and n8nOAuth2 protects the server with the built-in OAuth 2.1 server. Anything other than None means the client has to send credentials on every call.

  2. 02

    Create the credential once

    A credential in n8n is a saved set of secrets, created from the Credentials menu and reused across every workflow on the instance. For Bearer Auth and Header Auth, the setup is the same as the HTTP request credentials documented by n8n. Create it once, select it in the node, and it stays available for the next MCP server you build.

  3. 03

    Test before you publish

    The node shows two MCP URLs at the top of its panel, test and production, and a toggle between them. n8n registers the test URL when you select Listen for Test Event or Execute workflow on an inactive workflow, and the data lands in the canvas. The production URL is registered when you publish the workflow; its data shows up in the Executions tab instead.

Connections

What do you plug into the node?

n8n splits its AI nodes in two: a root node sits in the workflow, and sub-nodes hang off its ports, one type of sub-node per port. A sub-node never runs on its own. Here the root node is the trigger itself, and the only port is the one that carries the tools.

Ports (what plugs in)

  • Toolai_tool
01ToolThis port holds everything the client is allowed to call. The node lists the attached tools when a client asks, then runs the one it names, which is why nothing else connects to this trigger.

This port holds everything the client is allowed to call. The node lists the attached tools when a client asks, then runs the one it names, which is why nothing else connects to this trigger.

Key parameters

  • Optional: the node still validates without a tool, but a client then finds an empty catalog.
  • AI Agent Tool: hands a whole agent to the client as a single callable tool.
  • MCP Client Tool: chains an external MCP server into the one you publish.
  • Calculator and Think Tool: small utilities, useful to test the connection before wiring anything serious.
Use cases
a client that answers with live data usually gets two or three narrow tools rather than one that tries to cover everything.
Parameters

Which parameters configure the server?

The MCP Server Trigger node has 2 parameters. For each one: the node as you configure it in n8n, what the parameter changes, and our field notes.

01

Authentication

authentication

What you see in n8n

Notes & use cases

Sets the way clients authenticate against your MCP URL. It is the only thing between a public URL and your tools, so it deserves a decision.

Key parameters

  • None (none): no check at all, fine on a local test, risky once the workflow is published.
  • Bearer Auth (bearerAuth): the client sends a token, which is what the Claude Desktop gateway configuration expects.
  • Header Auth (headerAuth): the secret travels in a header you name, handy when a client cannot send a bearer token.
  • n8n OAuth2 (n8nOAuth2): protects the server with the built-in OAuth 2.1 server.
Use cases
a server that reaches a CRM runs on Bearer Auth from the first test, not from launch day.
02

Path

path

What you see in n8n

Notes & use cases

Defines the base path of the MCP URL, the part that comes after your n8n domain. n8n fills it with a random value so two MCP Server Trigger nodes never collide on the same instance.

Key parameters

  • Path: a text field, placeholder webhook, that you can set by hand and that accepts route parameters.
Use cases
a fixed path is what you want when n8n prototypes an API and the endpoint URLs have to stay stable across changes. Careful though: rewriting the path invalidates every client already configured with the old one, so change it before you hand the URL out, not after.
Need help

Need help automating MCP Server Trigger with n8n?

A person reads every message.

FAQ

MCP Server Trigger and n8n, the usual questions

01Is the MCP Server Trigger node included in n8n?
Yes. It belongs to the n8n AI package, shipped with n8n, so there is nothing to install and nothing extra to pay on the n8n side. It works the same on n8n Cloud, the hosted offer run by n8n, and on a self-hosted instance in Community Edition under the Sustainable Use license. A workflow built on one runs on the other. What can cost money sits elsewhere: if a tool you expose calls a model provider, that provider bills its own API calls under its own terms, and n8n adds nothing on top.
02What do you need for the n8n MCP Server Trigger node to work?
A path and an authentication decision, and that is the whole setup. Authentication accepts None, Bearer Auth, Header Auth or n8n OAuth2; the last three need a credential, which is a set of secrets saved once in the Credentials menu and reused everywhere. Bearer Auth and Header Auth follow the same setup as the n8n HTTP request credentials. You also need at least one tool attached to the Tool port, otherwise a client connects and finds nothing to call. A node with no tool attached is a server with an empty catalog, which answers but serves no purpose.
03What are the limits of the MCP Server Trigger node?
Three, all documented. Transport first: the node speaks Server-Sent Events and streamable HTTP, and it does not support standard input and output transport. Scaling next: in queue mode a single webhook replica is fine, but with several replicas every request under the MCP path has to reach one dedicated replica, or connections break. Reverse proxy last: nginx needs proxy buffering off on that location, and usually gzip and chunked transfer encoding off too. One more oddity, claude.ai custom connectors ask for an n8n sign-in even when Authentication is None.
04What should you attach to it first?
Start with one tool, not five. The Tool port takes sub-nodes such as AI Agent Tool, MCP Client Tool, Calculator, Code Tool, Think Tool, Wikipedia, SearXNG, Wolfram Alpha or the Vector Store Question Answer Tool. A Calculator is enough to check that the client connects, lists the tool and gets an answer back. Once that round trip works, swap it for the tool you actually care about. To hand an entire workflow to a client, attach it with the Custom n8n Workflow Tool node.
05n8n or Make to publish an MCP server?
It depends on where you want the server to live. n8n self-hosts through Docker or npm, or runs on n8n Cloud, so the endpoint and the data passing through it can stay on infrastructure you control. Make is hosted only, with no self-hosting option, and is billed per operation, which changes the cost model when a client calls the same tool all day. The visual logic differs too. Pick on hosting, data control and cost shape rather than on a feature list.
Hack'celeration Lab

Get our weekly integration tips.

No spam. Unsubscribe anytime.

Same platform

Other docs you might need on this stack.

Browse all integrations