n8n

What n8n actually meters

A production execution. n8n's docs define it as a single run of a workflow, and the count doesn't move with complexity: "It doesn't matter how many steps are in the workflow or how much data it processes."

The exclusions are where this gets genuinely cheap, and n8n documents them precisely:

• Manual executions from the editor don't count.

• Sub-workflow executions don't count. Only the top-level parent run is billed, so splitting one big workflow into a parent plus six children still costs one execution.

• Error workflow executions don't count, on any plan, since n8n 2.38.

• Polls that return no data don't count, so a trigger checking every minute and finding nothing costs nothing.

• Malformed webhook requests that fail before the workflow starts don't count.

What does count: every Schedule Trigger firing, whatever the outcome, and every inbound webhook request that starts a workflow, including one with an empty body.

Concurrency is metered separately. Starter runs 5 concurrent production executions, Pro 20 and Enterprise 200 or more. Excess executions queue in FIFO order rather than failing.

How credits work

n8n runs two credit balances, and they pay for different things.

Gateway credits cover AI models and tool services, so you can call OpenAI, Anthropic, Gemini, Brave Search or Firecrawl without your own provider account. n8n bills per request at the rates on its service pricing page, deducting from one prepaid balance shared across the instance. Top-up credits expire 12 months after purchase, and n8n spends free credits and soonest-expiring credits first. Credits aren't refundable and you forfeit them if you close the account. Auto top-up refills to a target balance, with an optional monthly cap. Gateway credits are Cloud only, on Starter and Pro.

AI Assistant credits pay for the workflow builder instead: 2,300 a month on Starter and 5,700 or 13,700 on Pro. They refresh monthly, don't roll over, and you can't buy more, so a heavy Assistant user upgrades the plan or nothing. Self-hosted Assistant use is bring-your-own-key with no credit limit.

What happens when you hit the limit

Workflows keep running. n8n doesn't block at the quota line, so an overrun never breaks an automation. Overage charges may apply if you don't move up a tier, and n8n's sales team reaches out as you approach the limit. Where nobody responds, overages get invoiced automatically 45 days after they occur.

Business is the one tier with a published overage rate: 4,000 EUR for each extra bucket of 300,000 executions. Per execution that's 0.0133 EUR, under the 0.02 EUR that Business's 40,000 entry bucket costs on monthly billing but over the 0.0105 EUR at its 500,000 rung, so overage only beats the plan at the bottom of the ladder.

Gateway credits behave differently, because they're prepaid. At zero balance, nodes using Gateway credits stop and nodes using your own API keys carry on. In-flight requests finish, the balance dips below zero, and n8n takes the difference out of your next top-up.