Skip to Content
Deploy & ConfigureMCP ServersSetup Examplesn8n

n8n MCP Server

This guide connects an n8n  workflow to auxilia as an MCP server. Once connected, an agent can trigger your n8n workflows the same way it calls any other tool.

Why connect n8n?

n8n already integrates with hundreds of apps (Google Workspace, HubSpot, Stripe, internal HTTP APIs, databases, and more). Rather than building a custom MCP server for each of those, you can:

  • Expose existing automations as agent tools — wrap a multi-step n8n workflow (“create a Jira ticket, notify the owner in Slack, log it to a sheet”) behind a single tool the agent can call
  • Reuse logic you already trust — the workflow keeps running in n8n, with its own credentials, retries, and error handling; the agent just decides when to fire it
  • Reach systems that don’t have an MCP server yet — if n8n can talk to it, your agent can too

This is the inverse of the n8n & Cloud Scheduler tutorial, where n8n calls auxilia. Here, the agent calls n8n.

How it works

n8n’s MCP Server Trigger node turns a workflow into a remote MCP server. Each Tool sub-node you attach to it becomes a tool the agent can discover and call. n8n publishes a stable URL for that server, which you register in auxilia as a custom MCP server.

┌──────────┐   tools/call    ┌────────────────────┐   runs    ┌──────────────┐
│ auxilia  │ ──────────────▶ │ n8n MCP Server      │ ────────▶ │ your apps    │
│ agent    │                 │ Trigger workflow    │           │ (Slack, CRM) │
└──────────┘                 └────────────────────┘           └──────────────┘

1. Build the MCP server workflow in n8n

  1. In n8n, create a new workflow
  2. Add an MCP Server Trigger node as the entry point
  3. Attach one or more Tool sub-nodes to the trigger — each becomes a tool exposed to the agent. Give every tool a clear name and description; the agent picks tools based on these, so write them the way you’d describe the action to a teammate.
  4. Build the rest of the workflow downstream (the actual automation)

2. Choose the transport and authentication

In the MCP Server Trigger node:

  • Transport — choose HTTP Streamable. auxilia speaks the Streamable HTTP  transport only; the older SSE transport is not supported.
  • Authentication — choose Bearer Auth (recommended). n8n lets you set a bearer token credential; auxilia will send it on every request. You can also use None for a quick test, but never leave a write-capable workflow unauthenticated.

Copy two things from the node:

  • The Production URL (not the Test URL — the test URL only works while you have the n8n editor open with “Listen” active)
  • The bearer token you configured

3. Activate the workflow

Toggle the workflow to Active in n8n. The Production MCP URL only responds when the workflow is active.

4. Register the server in auxilia

In auxilia, navigate to MCP Servers (admin only):

  1. Click Add MCP Server

  2. Fill in:

    • Name — e.g. n8n Automations
    • URL — the Production MCP URL from the trigger node
    • Authentication — API Key, and paste the bearer token

    See Authentication methods for how the key is encrypted and sent.

  3. Save

Because n8n uses a shared bearer token (not per-user OAuth), there’s no Connect step — the workspace shares one key. auxilia fetches the tool list from your MCP Server Trigger node immediately.

5. Add it to an agent

  1. Open an agent’s configuration page
  2. Click Add MCP Server and pick n8n Automations
  3. Review the discovered tools and set tool settings. Workflows that write to other systems are good candidates for needs approval until you trust the agent’s judgment.

If you later add or rename Tool sub-nodes in n8n, click Sync tools on the agent-server binding so auxilia picks up the changes.

Troubleshooting

No tools appear after adding the server

  • Confirm the workflow is Active and you used the Production URL, not the Test URL
  • Confirm the transport is HTTP Streamable, not SSE
  • Confirm each Tool sub-node has a name and description

”Unauthorized” or 401

The bearer token in auxilia must match the one configured on the MCP Server Trigger node. Re-copy it and re-save the server.

Intermittent failures behind a load balancer

If you self-host n8n in queue mode with multiple webhook replicas, MCP connections must be routed to a single dedicated webhook replica — Streamable HTTP relies on persistent connections to one instance. A single webhook replica works without any extra routing.

Sources: MCP Server Trigger node — n8n Docs 

BigQueryOverview