n8n

The connector works with your own n8n instance, on n8n Cloud or self-hosted, through n8n's public API and one API key. It lists workflows, reads their nodes and settings, activates or deactivates them, and reads every execution with its status and, when you ask for it, the data each node produced. It does not start a run on demand.

Is there an MCP server for n8n?

Yes. n8n ships a built-in MCP server in every instance. An owner or admin turns it on under Settings, then Instance-level MCP, and the page shows a server URL ending in /mcp-server/http. Clients such as Claude and ChatGPT connect to it with OAuth or with an API key, and n8n lists the connected clients so access can be revoked. Through it a client can search workflows, run the ones you have enabled for MCP and, from n8n 2.13.0, create and edit workflows and data tables. Tulina's connector is the other route, held in the same MCP link as the rest of your tools.

Are MCP and n8n the same?

No. MCP is the protocol an AI client uses to call tools. n8n is an automation platform that can sit on either side of it: its instance-level server exposes your workflows to an AI client, and its MCP Server Trigger node exposes the tools of a single workflow. The Tulina connector does neither. It reads and manages your instance from the outside, so an assistant can answer what ran and what failed.

How do you connect n8n to Claude and ChatGPT?

Through Tulina's MCP server, in three steps:

1. In n8n, open Settings, then n8n API, and create an API key. Copy your instance URL too, for example https://acme.app.n8n.cloud or your self-hosted address.
2. Paste both into Tulina's n8n connector: the key and the instance URL. The key is stored in the vault and never shown to the model. Several instances can be connected, each under its own name.
3. Add https://mcp.tulina.ai/mcp to Claude or ChatGPT and sign in with your Tulina account. The MCP page has the same link.

Tulina vs n8n's official MCP server

n8n's server is the first-party route and runs inside your instance. Tulina's reaches the instance through the public API and sits beside your other tools. The vendor column below comes from n8n's documentation page, "Connect to n8n MCP server", read in October 2026.

If the model should run a workflow or build one, n8n's own server is the better choice. Tulina fits when the n8n question, what ran and what failed, belongs to a longer piece of work that also touches your inbox, your tickets or your messaging.

What you can ask it

"Which workflows are active right now, and which of them failed overnight?" The agent lists active workflows, then lists executions with the error status and groups them by workflow.

"Why did the lead-sync run fail at 9:40?" The agent finds the execution, reads it with its per-node data included, and points at the node that errored and the input it received.

"Pause the invoice reminder workflow until Monday." The agent finds the workflow by name, confirms which one it is and deactivates it. Reactivating it is the same call the other way.

Teams that run n8n next to another automation tool can pair it with Zapier or Make, and send failure summaries to Slack.

What the n8n connector can't do

• It does not run a workflow. The tools are list, read, activate, deactivate and the execution reads. Starting a run is done in n8n, or through n8n's own MCP server.
• It does not create or edit workflows, credentials or tags. It lists tags and reads workflows.
• Per-node execution data can be large. The agent asks for it on one execution at a time, not across a list.
• Deactivating a workflow stops its triggers and schedules at once. Ask the agent to name the workflow and confirm before it does.
• No tool walks through several n8n instances at once. A total across instances is one call per instance.