Getting started

Credits & pricing

Every call is priced in credits: some tools cost a flat amount, others cost whatever the model costs. Here is the full table, how batches are billed, and how to keep spend predictable.

How pricing works

There is one currency — credits — and two ways a call is priced. Flat-rate tools always cost the same amount. Model-priced tools cost whatever the model you selected costs, which is the costPerReq value returned by list_models. When a model has no cost recorded, the server bills 15 credits.

Cost per tool

ToolEndpointCostDetail
get_user_infoGET/auth/user-dataFreeThis endpoint never deducts credits.
list_modelsGET/playground/modelsFreeThis endpoint never deducts credits.
generate_imagePOST/playground/generateModel costDeducts the selected model's `costPerReq` (15 credits when the model has no cost recorded).
generate_transparent_imagePOST/playground/generate-transparentModel cost + 8Generation is billed at the model's rate (15 credits by default) plus 8 credits for the cutout pass.
edit_imagePOST/playground/editModel costBilled at the model's rate. With `action: "remove_bg"` the call costs 8 credits and with `action: "upscale"` it costs 15.
remove_backgroundPOST/playground/remove-background8 creditsFlat 8 credits per call, regardless of image size.
upscale_imagePOST/playground/upscale15 creditsFlat 15 credits per call at any scale factor.
text_to_svgPOST/playground/text-to-svg5 creditsFlat 5 credits per call.
compress_imagePOST/playground/compress1 creditFlat 1 credit per call. No model is involved, so it is fast and cheap.
convert_formatPOST/playground/convert-format1 creditFlat 1 credit per call.
multicallPOST/playground/multicallSum of the batchEvery item is priced by its own tool and the total is deducted once, up front, before the batch runs.

Batches

multicall prices each item with its own tool's rules, sums them, and deducts the total once before the batch runs. A batch of three generations at 15 credits plus one background removal at 8 costs 53 credits, taken up front.

multicall response (trimmed)
{
  "success": true,
  "count": 3,
  "totalLatency": "11.07s",
  "deductedCredits": 43,
  "userCredits": 2374
}

When credits are taken

The balance is checked and the credits are deducted before the request is handed to the model. That ordering is what stops a batch from running away with a balance it does not have — but it also means a call that fails downstream has already been charged.

Failed calls are not auto-refunded

Credits are deducted up front and are not returned automatically when a generation fails after that point. If a run fails repeatedly, stop retrying and contact support rather than burning the balance.

Plans

Credits arrive with your plan and reset monthly. Top-up packs are available if you run out mid-cycle.

Free$0/month

30 credits / month

Starter$20/month

1,000 credits / month

Pro$50/month

3,000 credits / month

Enterprise$100/month

8,000 credits / month

Full plan comparison and credit packs live on the pricing page.

Keeping spend down

  • Check user.credits with get_user_info before starting a long agent run.
  • Read userCredits from each response instead of polling the balance separately.
  • Prototype prompts on a cheaper model, then switch models once the prompt is settled.
  • Use compress_image (1 credit) rather than regenerating at a smaller size.
  • Batch with multicall — it does not reduce credits, but it removes the round-trips that make agents retry.

FAQ

How much does one image cost in credits?

Generation is priced by model — the call deducts that model's costPerReq, or 15 credits when no cost is recorded for it. Flat-rate tools are cheaper and fixed: background removal is 8, upscaling 15, text-to-SVG 5, compression and format conversion 1 each.

When are credits deducted?

Before the request reaches the model. The balance check and the deduction happen up front, so a call that fails after that point does not return its credits automatically — contact support if a run fails badly.

Do checking credits or listing models cost anything?

No. get_user_info and list_models are free to call as often as you like.

How is a multicall batch billed?

Each item is priced by its own tool, the total is summed, and that total is deducted once before the batch starts running.