Skip to main content
WorkflowsCredits & Limits

Credits & Limits

This page has been cleared and is ready for updated Builder Studio documentation.

Running a workflow consumes credits. Credits are owned by the workspace, not by an individual canvas or user, so every run in a workspace draws from the same shared balance. Only the executable steps in a run cost credits — layout and content-only nodes are free.

Where credits live
Credits are a workspace-scoped balance with a transaction log. You can see the current balance in the canvas controls and in workspace billing settings, where the transaction history shows each deduction, refund, and purchase.

What a run costs

Each executable node type has a fixed per-action credit cost. A run's total is the sum of the costs of the executable steps in its subgraph. Nodes with no executor (text, shapes, documents, code) and passthrough media nodes contribute nothing.

ActionCredits per run
Text Generation1
Image Generation21
Text-to-Speech21
Speech-to-Speech21
Sound Effects21
Video Generation621
Costs are per action, not per provider option
The numbers above are fixed per-action costs set with a margin over the underlying provider cost. They do not vary by model, resolution, or duration within an action. Node types that do not appear in this table (for example media passthrough, prompt, and webhook nodes) add nothing to the total. These figures can change as vendor pricing changes, so treat the live balance and transaction log in your workspace as the source of truth.
Cost of a sample runtext
Run a graph: Prompt -> Text Generation -> Image Generation Prompt              0   credits  (no executor cost)Text Generation     1   creditImage Generation    21  credits----------------------------------Total charged       22  credits   (sum of executable steps only)

Starting balance

New accounts receive a one-time starting grant of 2,000 credits, applied to the owner's oldest workspace on first use. Additional workspaces start at zero unless they are topped up. You can buy more credits from the billing area; purchases are added to the workspace balance and recorded in the transaction log.

When you are charged

Credits are deducted up front, when a run is admitted, never twice for the same run:

  • Manual studio runsare charged at admission against the workspace balance. Before accepting the run, Builder Studio checks that the balance covers the run's total; if it does not, the run is rejected with an insufficient-credits error.
  • Server-triggered runs — API, webhook, MCP, user-MCP, and replay — are charged by the dispatch step that runs the graph, again with an up-front balance check.
  • Charges are idempotent per execution: replaying or retrying the same execution id does not double-charge.
Insufficient credits stops the run
If a workspace cannot afford a run, the run is not admitted. For server-triggered runs this surfaces as an aborted, insufficient-credits result rather than a partial execution.

Refunds on failure

Credits charged for a run are restored when the run ends in error or cancelled. The refund is verified against the original deduction for that execution, is applied once, and is recorded as an execution-refund transaction. Successful runs are not refunded. See Running workflows for the status model.

Watching your usage

  • The workspace credit balance is shown in the canvas controls so you can see it while you work.
  • Workspace billing settings show the balance, a low-balance warning when it runs down, and the full transaction history of deductions, refunds, and purchases.
  • Workspace analytics summarize execution activity and credit spend over time.

Rate limits

Separate from credits, runs are subject to per-minute rate limits so a single caller cannot overwhelm the system. These caps are independent of your credit balance — hitting a rate limit returns a retry-after response and does not consume credits. Limits apply to triggering, streaming, and reading executions, and to the underlying generation endpoints. Verified examples include:

SurfaceLimit (per minute)
Manual browser start60
Execute API (server-to-server)30
Webhook trigger60
Image / video / audio generation30 each
Text generation60
Execution status reads120
Replay / redrive10 each
Cancel30
Rate limits can change
These per-minute caps are operational defaults and may be tuned over time. When you are rate-limited, back off and retry after the indicated delay.
Related
See Workflows overview for what counts as an executable step, Running workflows for trigger types and status, and Billing & usage export for exporting usage data.

Was this page helpful?