text
1 credit = $0.0001 of provider cost
credits = tokens x price_per_1M x 10 / 1000Because a credit is derived from cost, the model you pick changes how far a balance goes rather than what a balance can cost. An AI step is charged from the token counts the provider actually returned; a run also carries one credit of overhead.
| Charge | Credits |
|---|---|
| Run overhead | 1 per run |
| HTTP request | 1 |
| Email, Slack, Sheets, app action | 2 |
| Logic, transform, code, triggers | 0 |
| AI step | Priced from the model's returned token counts |
Runs are refused with 402 once an allowance is spent, and workflow creation with 409 once the plan's automation cap is reached. Both responses carry an X-Nudge-Limit header naming which limit was hit: credits, runs, or workflows. Nothing is deleted when a limit is reached.