mrkeyoor.com_
Tue 01 Sept 16:17 UTC

This page is what happened when we ran it. The project itself — code, maintenance, community — has its own review: full repo review →

AutomationMCP servermeasured 19 Aug 2026

n8n

n8n-mcp is a community server by czlonkowski that hands your agent the n8n automation platform's node documentation. It searches nodes and workflow templates, fetches a node's full parameter reference, and validates a node config or a whole workflow JSON before you paste it into n8n. It talks to docs, not to your live n8n instance.

BootBOOTS ✓We started it and initialize came back in 15.0s.
Tools7tools/list payload 11,768 bytes
Context tax~2,942estimated tokens for the whole tools/list, chars/4 estimate
Resources / prompts2 / counts returned by resources/list and prompts/list
Measured19 Aug 2026one run of the harness, clean environment, no credentials · how we test

How we started it

npx -y n8n-mcp
Verdict

Use it if you write n8n workflows with an assistant and are tired of hallucinated node parameters, because validate_workflow catching a bad JSON before n8n does is worth the seven tool slots. The cost is the start: we booted it and initialize took 15,010 ms, which is the slowest kind of wait to hit at the top of a session, and it prints a telemetry notice on first run that you have to opt out of yourself. If you only touch n8n twice a year, skip it and read the node docs in a browser.

Use it if

You build n8n workflows regularly and want the agent to check node parameter names against the real reference instead of guessing them
You generate workflow JSON with an LLM and want validate_workflow to reject it before n8n silently accepts a broken node
You want template search inside the editor loop, so the agent can start from an existing community workflow instead of a blank canvas
You accept a slow cold start in exchange for a small, well-described toolset

Skip it if

You expected it to run your workflows: this server reads documentation and validates configs, it does not trigger or manage executions on your instance
You start fresh agent sessions all day and cannot afford a fifteen-second initialize each time
You are unwilling to send anonymous usage telemetry, or unwilling to run the opt-out command before you trust the process
You are on an old or self-patched n8n version, where the bundled node reference may describe parameters your instance does not have
Your context budget is already tight and n8n work is a small slice of what the agent does

The bill for your context window

ToolEst. tokensDescribed?Required params
search_templates~685yes (319 chars)0
validate_node~562yes (244 chars)2
get_node~533yes (399 chars)1
validate_workflow~508yes (128 chars)1
search_nodes~321yes (219 chars)1
tools_documentation~178yes (194 chars)0
get_template~153yes (133 chars)1

Seven tools cost 11,768 bytes of tools/list, which the harness estimates at 2,942 tokens using chars/4. That is more than the median server in the same batch and more than seven tools usually justify, and the reason is that the descriptions carry worked examples rather than one-liners: search_templates alone is estimated at 685 tokens, and validate_node at 562. It is money reasonably spent. These are tools whose whole value is the agent knowing the exact argument shape, and Notion in the same batch spends several times as much on a toolset that is not several times more useful.

Setup reality

Nothing to configure. We ran it as npx -y n8n-mcp with a clean environment and no credentials, and the harness reported no environment variables were required for boot. Transport is stdio, no OAuth, no API key, no running n8n instance needed, because the node reference ships inside the package. First run pays the npx download plus a database warm-up, and it prints a telemetry banner with an anonymous ID; the opt-out is a separate npx telemetry disable command you run once. It also exposes two resources alongside the tools.

Questions people ask

Can n8n-mcp run or edit workflows on my n8n instance?

Not in what we measured. The seven tools we saw search nodes and templates, fetch node reference, and validate node or workflow JSON. Nothing in that list connects to a live instance or triggers an execution, and the server never asked us for a host or API key.

Why does it sit there before it answers?

Our initialize took 15,010 ms on a cold npx run. Part of that is downloading the package, and the server also loads its bundled node database before answering. Pin the install locally rather than resolving through npx if that wait lands in every session.

Does it phone home?

Yes, and it says so. On first run it printed a notice about anonymous usage telemetry with a generated ID, listing error patterns and timing as collected and credentials and workflow content as not collected. Disabling it is a separate command, not a flag.

Is it worth the tokens for occasional n8n work?

Probably not. At roughly three thousand estimated tokens it sits above the batch median, and the payoff only shows up when you are writing node configs often enough that a wrong parameter name costs you real time. Occasional users get more from the docs site.

How this page is made: the server is spawned as an unprivileged user with a clean environment and no credentials, then asked for its tools, resources and prompts over stdio. Token figures are estimates at four characters per token, not a tokenizer count. One run, one machine. Corrections: contact the desk.

← All measured MCP servers