What it costs to keep an automation running, per month
The build price is the number everyone quotes. This is the other one, the one that arrives every month afterwards and decides whether a working system stays switched on. Move the numbers to match your case.
You pay these providers directly, on your own accounts. We add no markup and never have.
This page stays in dollars even with the site switched to pounds or euros. OpenAI, Anthropic, n8n and the telephony carriers all bill in dollars, so that is what your statement will say. Putting a converted figure here would print a number no invoice of yours ever carries.
| n8n Cloud, 3,000 runsn8n.io/pricing, plus our own reading of the tier ladder above Starter | $41.18 |
| Model tokens, 3,000 runsour own assumption from typical payloads | $37.50 |
$944 a year
Where these numbers come from
Each line names its source under the label. The platform, telephony and number rates are vendor list prices read in August 2026. The voice model rate is derived from OpenAI's published audio token pricing at a 60% talk-time ratio, which is our assumption and the first thing you should check against your own recordings.
Two rows are marked as our own assumptions rather than published rates: the per-run token cost, which comes from typical payload sizes we see rather than from a price list, and the hosting line. Both move a lot with what you actually build.
What this deliberately excludes: failed and abandoned calls that still consume telephony, recording and transcript storage, and US outbound compliance registration. Those are real and they are covered in the running costs post and, for voice specifically, in the voice cost breakdown, where every vendor figure carries a read-live status.
Vendor prices move. This page is an estimate you can check, not a quote. The exact stack for your build and its monthly cost are written into every quote we send, before you sign anything. Build prices are here.