Skip to main content
Overten uses a credit balance to meter API usage. Credits are consumed only when a generation completes successfully — if your request fails at any point before or during execution, you are not charged. This page describes exactly what costs credits, how the deduction is calculated, and what to do when your balance runs low.

​
What costs credits

Only successful generation runs consume credits:
  • POST /excel/generate
  • POST /word/generate
  • POST /slides/generate
  • POST /excel/analyze
  • POST /text/humanize, /text/rewrite, /text/new-content
Everything else is free:
  • GET /runs, GET /runs/{run_id}, GET /tasks/{task_id}
  • GET /verify, GET /webhooks
  • POST /files (file upload), DELETE /images/{id}
  • POST /preflight (preflight-only validation)
  • GET /runs/{run_id}/download and edit-URL rendering
  • File storage and bandwidth (not billed separately in v1)

​
How a run is priced

The credits_used field returned on every completed response is the authoritative charge for that run. It’s calculated from the actual model-token usage across all LLM turns inside the run, plus the agent’s tool activity — not from a fixed per-format rate. Prices depend on:
  • Which models were used — this varies by request shape and your account tier.
  • How much the agent had to iterate — a short prompt on a single sheet costs a fraction of a multi-sheet financial model that took 20 agent turns.
  • Whether the run resumed from state — follow-up turns on an existing run_id are cheaper because the agent doesn’t re-inspect the full document.
Overten doesn’t publish a static per-format rate card in v1. If you need budgeting numbers for procurement or capacity planning, contact sales@overten.ai for the pricing worksheet for your tier.

​
Balance check before a run

Before a generation request is accepted, Overten verifies that your balance is above a safety floor. If it isn’t: 
402 Payment Required
{
  "success": false,
  "error": "insufficient_credits",
  "message": "Insufficient credits. Required: 50, balance: 12."
}
The Required value is a conservative estimate — enough to complete the work. The actual deduction after a successful run is usually equal to or less than this estimate, because it reflects real usage rather than a worst-case projection. This check runs before preflight and before any agent work starts, so a 402 response never costs credits.

​
When you’re charged

ScenarioCredits charged?
Request rejected by balance check (402)No
Request rejected by preflight (preflight_failed)No
Request rejected by auth or permissionsNo
Agent runs and succeedsYes — see credits_used on the response
Agent runs and fails mid-executionNo (v1 behavior)
Sync call auto-promoted to async, then succeeds on the workerYes
Async worker job failsNo
In v1, runs that crash after starting are not charged. This is intentionally customer-friendly behavior. If this policy changes, we’ll update this page and the changelog before enforcing it.

​
Checking your balance and topping up

Your credit balance is visible at app.overten.ai → Settings → Billing. Top-ups are applied immediately — the next request sees the updated balance. Every completed response includes a credits_used field, so you can reconcile spend from your own logs without needing dashboard access.

​
Audit trail

Every credit movement — deduction, top-up, refund — is recorded against your org. If you need to investigate a specific charge, reference the run_id from the response payload when opening a support ticket. The Overten team can pull the matching record and walk you through the breakdown.

​
Refunds

If a completed run produced something unusable — wrong format, corrupted output, or an edge case that slipped through preflight — open a support ticket with the run_id. We’ll review the internal record and issue a credit adjustment on a case-by-case basis. Refunds are not issued for results that were generated as requested but didn’t match your expectations. If the agent produced an artifact for the prompt you submitted, the fix is a better prompt or a follow-up turn — not a refund. Work with support on patterns that misbehave and we’ll tune the agents.

​
Observability (enterprise)

Per-run cost attribution — which agent turn, which tool call, which model — is available to our team via our internal observability backend. File a support ticket with a run_id if you want a breakdown of why a specific run cost what it did. Customer-side observability export (your own Langfuse or Datadog destination) is available on the enterprise tier. Contact sales@overten.ai to set it up.