Air Pipe is a backend you describe in config: the file that runs your schedules and flows also publishes a documented HTTP API, Prometheus metrics and an OpenTelemetry trace for every request. Managed or self-hosted, with 3,000 requests a month free — n8n Cloud has no free tier.
Air Pipe endpoints are readable config — build them from a prompt, by hand, or visually, all producing the same thing the engine runs. Diff them, review them, deploy them like code.
Traces, metrics and OpenAPI docs are generated for every endpoint automatically — no extra nodes, no separate instrumentation to maintain.
n8n Cloud has no free tier; every plan starts as a trial. Their cheapest paid plan, Starter, is €20 a month billed annually for 2,500 workflow executions. Air Pipe's free tier is 3,000 requests a month, every month — more monthly volume than n8n's entry paid plan, for nothing. Past that, a request unit covers a request up to a second, so a five-minute polling loop that would burn thousands of executions barely registers here. (n8n figures from their pricing page on 30 July 2026, annual-billing toggle; monthly is around 20% higher. Compared per run, on sub-second workflows.)
n8n can start a workflow from a webhook or an incoming MQTT message. Air Pipe goes further — it is the realtime layer your app connects to: WebSocket and MQTT pub/sub with presence, history-rewind and retained messages, served to your users from the same config, with your pipeline running on each message. Realtime out to your users, not only a trigger coming in.
If a visual canvas is how your team wants to build and reason about automation, that is n8n's design centre and it self-hosts properly. Air Pipe suits you when the thing you want at the end is a service: an HTTP or MCP endpoint with auth, docs and traces, that happens to run on a schedule too.
Run the single cross-platform binary yourself, or use the managed cloud. Either way your data stays on infrastructure you control.