Claude + n8n: What the Integration Can (and Can't) Do in 2026
n8n has one of the most complete MCP stories of any tool in 2026 — there’s just no Anthropic directory connector to enable. Instead you get real primitives: a built-in instance-level MCP server Claude Desktop can log into, an MCP Server Trigger node that turns any workflow into an MCP endpoint Claude can call, an MCP Client Tool node that lets n8n’s own AI Agent call out to Claude-side servers, a hugely popular community MCP server for building workflows, and a plain REST API behind all of it.
What none of them change is the shape: Claude only talks to n8n inside a session you start. n8n’s own triggers — webhooks, schedules, new-row watches — still can’t wake Claude up.
Here’s each route, when to pick it, and what to use when you want n8n work that happens without you at the keyboard.
Route 1: n8n’s built-in instance MCP server
This is the headline. n8n now ships a native, instance-level MCP server so clients like Claude Desktop can connect directly to your n8n and work with it — no JSON copy-paste, no self-hosted bridge. See Connect to n8n MCP server in the docs.
Once connected, Claude can:
- Search and discover the workflows in your instance
- Build and edit workflows (from n8n v2.13 onward)
- Execute workflows you’ve marked as MCP-available, and create/manage data tables
The endpoint is https://<your-n8n-domain>/mcp-server/http, and it supports two auth modes: OAuth2 (recommended for coding agents) or a personal access token. You enable which workflows are exposed from the MCP settings page (added in n8n v2.2.0). Note that execute_workflow defaults to production mode — clients can test unpublished versions but can’t silently overwrite a published one. It works on Cloud and self-hosted instances.
Pick this when you want to say “look at my n8n instance and add a Slack notification step to my lead-intake workflow” and have Claude actually do it against the live editor.
Route 2: MCP Server Trigger vs. MCP Client Tool nodes
Two node-level primitives that people constantly mix up, because they point in opposite directions:
- MCP Server Trigger makes n8n the server. Drop it into a workflow, wire tool nodes underneath it, and n8n exposes a Streamable HTTP / SSE endpoint that any MCP client — Claude Desktop, Cursor, your own agent — can connect to and call. This is how you publish a curated set of n8n actions (say, “create Jira ticket,” “look up customer in Postgres”) as tools Claude can invoke mid-conversation. Unlike normal triggers, it only connects to tool nodes, not a downstream chain.
- MCP Client Tool makes n8n the client. Attach it to an n8n AI Agent node, point it at an external MCP server’s endpoint, and your n8n agent can discover and call that server’s tools during a run.
For the “Claude + n8n” question, the Server Trigger is the one that matters: it’s how you hand Claude a safe, named menu of n8n actions instead of API access to everything.
Route 3: the community n8n-mcp server (build workflows in Claude)
If your goal is authoring n8n workflows with Claude’s help, the community server czlonkowski/n8n-mcp (~22k stars) is the standard. It gives Claude structured documentation for 2,000+ n8n nodes plus a large template library and validation tools, so you can prompt Claude Desktop or Claude Code to design a complete, valid workflow node by node. Configure it with your n8n API credentials and it gains create, update, deploy, and execution-management tools too — Claude can build the workflow and push it to your instance.
It runs as a local MCP server you add to your Claude Desktop config (claude_desktop_config.json) or Claude Code. Because it’s node-documentation-aware, it’s noticeably better at producing workflows that actually validate than asking Claude to freehand n8n JSON.
Route 4: n8n’s public REST API + API key
Under everything sits n8n’s public REST API, which manages workflows, executions, credentials, tags, users, and projects. Auth is deliberately minimal: send your key in an X-N8N-API-KEY header — no OAuth flow, no HMAC. Create a key under Settings → n8n API, give it a label and an expiration (90 days is a sane default), and on Enterprise you can scope it to specific resources.
You don’t call this by hand from Claude, but it’s the plumbing the MCP routes above sit on, and it’s exactly what a hosted assistant uses to reach n8n with your own key.
The limits that actually matter
- No Anthropic directory connector. n8n isn’t a one-click connector in Claude’s directory — you set up an MCP server (built-in, node, or community) or use the API.
- No triggers. Every MCP route runs inside a session you start. An n8n webhook firing at 2 a.m. executes its workflow as always, but Claude takes no part in that run — nothing on n8n’s side opens a Claude conversation.
- Laptop-bound. The community server and Claude Desktop config are local. Close the laptop and the scheduled thing you imagined doesn’t happen.
- claude.ai on the web can’t reach a local/desktop MCP setup; the built-in instance server needs a reachable n8n URL and auth.
Claude + n8n is excellent for building, editing, and running n8n workflows on demand while you’re in the chat. It’s not built for being the always-on brain that reacts to events itself.
If you want n8n work that runs on its own: Carly
Carly is an AI assistant that reacts to events itself and runs in the cloud — the always-on half that Claude-plus-n8n doesn’t cover.
- When a new lead lands in your CRM, Carly enriches it, drafts the intro email, and sends it — no chat window open.
- Every Friday at 8 a.m., Carly pulls the week’s unpaid invoices and emails you a summary.
- When a customer replies to a thread, Carly reads it, updates the deal, and books the follow-up.
No-code setup. You describe the job in plain English and Carly interviews you about the details, then assembles the workflow with you — no canvas, no node wiring. Actually sends — it delivers through Gmail and Outlook, attachments included, rather than handing you output to relay. Connects to anything — 200+ native integrations plus any tool with an API via your own key. Carly reaches n8n through your own n8n API key (paste it on the integrations page), so it can drive your workflows on its own triggers.
AI agents start at $35/month, and steps in a workflow that don’t use AI run free and unlimited.
n8n itself is a fine tool — it has its own in-editor AI Assistant and AI Agent nodes if you want to build and host automations yourself. The honest contrast is hosting and shape: n8n is a canvas you run and maintain; Carly runs and reasons on its own, hosted, with no self-hosting.
Claude vs Carly
| Claude (+ n8n MCP) | Carly | |
|---|---|---|
| Read/search your n8n workflows | Yes | Yes |
| Build/edit an n8n workflow in a session | Yes | Builds its own flows for you |
| Run a workflow when you ask in chat | Yes | Yes |
| Acts on triggers and schedules | No | Yes |
| Works while your laptop is closed | No | Yes (cloud) |
| Sends email itself | No | Yes — Gmail + Outlook |
| Setup | MCP server config / API key | Interview, no code |
| Pricing | Pro $20 / Max $100–$200 | AI agents from $35/mo |
Claude is the best co-builder for n8n workflows; Carly is the one that runs the work on its own.
Frequently Asked Questions
Does Claude work with n8n?
Yes. There’s no Anthropic directory connector, but n8n ships a built-in instance-level MCP server Claude Desktop can connect to, plus MCP Server Trigger and Client nodes and a REST API. The popular community server czlonkowski/n8n-mcp also lets Claude build workflows against your instance.
Is there an official n8n MCP server?
Yes — two ways. n8n has a native instance-level MCP server at https://<your-domain>/mcp-server/http (OAuth2 or access token) for connecting a client like Claude, and the MCP Server Trigger node turns any single workflow into its own MCP endpoint. For authoring, most people use the community n8n-mcp server.
Can claude.ai on the web access n8n?
Not for local or desktop-only setups. The community server and Claude Desktop config run on your machine. The built-in instance MCP server can work with a reachable n8n URL and valid auth, but there’s no in-chat toggle on claude.ai the way a directory connector would provide.
Can Claude run my n8n workflows automatically?
Only when you ask it to in a session. Claude can execute an MCP-exposed workflow on demand, but it can’t react to an n8n trigger on its own. For workflows that fire on events and schedules while you’re away, that’s what an agent like Carly is for — starting at $35/month.
More: Claude connectors · Claude vs Carly · Can Claude send emails · Claude + Make · Claude + Zapier · n8n alternatives · n8n vs Zapier
Ready to automate your busywork?
Carly schedules, researches, and briefs you—so you can focus on what matters.
See what people say
"Before Carly, I relied on a Calendly link, but the whole process felt impersonal and not very professional. Carly changed that by handling all the back-and-forth, so I'm no longer stuck in endless email threads trying to line up schedules.
Now Carly reaches out to candidates, shares my real-time availability, lets them pick a slot, then sends a Zoom link and drops it straight into my calendar. She sends reminders to both of us before each call, which has significantly reduced no-shows and last-minute confusion.
On top of scheduling, Carly acts like a full executive assistant, sending me my schedule the night before so I can prepare for each call. It reminds me of the old x.ai assistant, but Carly is noticeably smarter, faster, and better suited to my healthcare recruitment business."
