Skip to main content
Credits meter what your workspace consumes. Both surfaces are read-only: credits are granted, and spending is recorded.

​
Balance

The balance reports three numbers: Credits are held per workspace, not pooled across an organization. A team that runs a lot draws down its own balance.
Every human-facing field is expressed in credits. No surface converts credits to a currency amount.

​
Ledger

Entries are newest first. Each carries the amount, a reason, when it happened, and an action category — build or execution — so you can tell package builds apart from runs. The ledger is the answer to “where did the credits go”. A balance that dropped faster than expected almost always has a visible cause here.

​
What consumes credits

Two things: building a package, and executing a run. The usual surprises: A schedule firing more often than intended. A */5 * * * * expression left over from testing runs 288 times a day.
A retry loop. A MUST_HAPPEN step that fails retries with backoff before giving up, doing that work repeatedly on every run.
An agent staged live that should be dry-running. Dry Run still executes, so it still consumes — but it’s the right mode while you’re iterating on a spec, and it won’t compound with real side effects.

​
Keeping it predictable

Check schedules after testing, clear the ones you don’t need, and look at failed stories periodically rather than only when something breaks.

​
See also