Usage
Understand ActionFlow credit estimates, admission holds, settled usage, run-level records, and the limits that are not enforced by the current storage path.
Usage
Usage is the credit activity associated with one ActionFlow. Open it from Actionflow Studio or ActionFlow → Usage (/actionflows/{actionFlowId}/usage).
Estimate versus settled usage
Actionflow Studio can show an estimate before a run when the cost-calculation feature is enabled for the organization. The estimate is calculated from the current graph/model configuration and is not a bill.
The runtime uses a different lifecycle:
- It checks the organization's available credit balance and reserves a run hold at admission.
- It executes the graph and calculates usage from executed node occurrences and returned AI usage.
- It settles the hold with the calculated amount. A failure can retain billable usage from the prefix; a run with no billable usage can release the hold.
- The settled amount is shown in run history and Flow Usage.
The estimate can differ from the settled amount. Provider usage, fallback modality minimums, loops, failed nodes, and the model's returned usage all affect the final calculation. Organization-owned BYOK AI usage is handled differently from platform AI usage, so do not infer a charge solely from the selected model label.
The current credit path does enforce the organization's available credit balance and run reservation. Do not extend that statement to every CMS plan field: file-size and organization-storage values may be configured or calculated without an upload-time quota check.
What the Flow Usage view shows
The Flow Usage table reads credit-usage records for the selected ActionFlow. Depending on the record, the description identifies the run or billable usage line, including service/model metadata when it is available. Run history groups the settled amount for an execution and can show the per-node credit breakdown.
Use the following surfaces together:
- Actionflow Studio estimate: expected cost before a run.
- Live Run Output: progress and current credit metadata while the run is active.
- Run History: settled run amount, executed nodes, failure context, and retry links.
- Flow Usage: organization credit records filtered to this ActionFlow.
- Organization → Credits: organization balance and all-flow/agent activity.
Admission and failure cases
A run can be rejected before execution if the organization has no available credits. A run can also fail after a prefix has consumed billable usage; the history row retains the calculated prefix usage. Cancelled or failed runs with no billable usage release the reservation. These are runtime credit behaviors, not estimates.
Try mode has no organization Usage panel and does not bill an organization.