DocumentationIntegrations

n8n

The n8n connector makes the workflows of your own n8n instance callable in the anymize chat, via access token and without a detour through a third-party service.

3 Min

What you can do with it

The assistant finds your workflows, runs them and can edit them. Which workflows are visible at all you decide in n8n: only what you enable there for remote access appears in chat.

  • “Which workflows can I start from here?”
  • “Start the workflow for invoice import.”
  • “What did the last run of this workflow produce?”

How to create the n8n access token

  1. 01In n8n open Settings, then Instance-level MCP. Owner or admin must enable MCP access once.
  2. 02Under Connection details click Connect and choose the API key or Access Token tab.
  3. 03Copy the access token immediately. Next time you visit it is only shown shortened.
  4. 04Make sure the workflows that should appear in chat are enabled for MCP in n8n.

Official MCP guide: https://docs.n8n.io/connect/connect-to-n8n-mcp-server . That is a different place from Settings, n8n API for the classic REST API.

How to connect n8n

  1. 01In anymize open Apps, then Connectors, and search for n8n.
  2. 02Enter the base address of your n8n instance, for example https://n8n.your-company.com. anymize adds the rest of the path itself.
  3. 03Paste the access token from the section above, test the connection and connect.

Does this work with my self-hosted n8n?

Yes, that is even the normal case. anymize talks to your instance wherever it sits. There is one condition: the address must be reachable over https. Plain http addresses and addresses from a private network that is not reachable from outside are rejected.

That is not harassment, but the boundary that stops a connection pointing into a network where it has no business.

Does the AI run workflows on its own?

No, it only runs what you ask for. But because a workflow run has real effect and the connector can also edit workflows, the tile says “read and write”.

n8n runs with direct access, so without the in-between privacy agent. The main model sees the real answers from your workflows. That is also stated in the connect dialog.

What is anonymized

What you write yourself in chat goes through anonymization: customer names, addresses, document numbers and credentials in your request are replaced by placeholders before the query goes to an AI model.