Do the Sume Agent and API video draw from one credit balance?

Yes. Sume keeps one credit pool, and generation in the app, the Agent, Formats and API calls all spend from it. A top-up or a plan's included usage is available to every surface, and there is no separate Agent wallet. Included usage figures come from Sume's plan catalog; the Billing and subscription page in the dashboard shows what is left on your workspace.

One pool, several surfaces

The billing docs describe credits as a single pool across the products. The order in which lots are spent is the same everywhere: included usage, then top-ups, then grants.

When the pool runs dry

An API job that the balance cannot cover returns 402 insufficient_credits before any work starts. A queue that is full returns 429 queue_full instead. For video jobs, Sume reserves the estimated cost at submit and captures it on completion, or releases it if the job fails.

Because the pool is shared, heavy Agent use can leave less for an API batch. If you run both, check the balance before a large submit, and treat top-ups as the shared buffer. See how top-ups work through the API.

A practical habit

Read the balance and usage endpoints at the start of a batch, and compare the estimated total with the balance minus any reserved amount. The reserve post explains why the balance can drop before a queued job begins. A team that wants a firm split between Agent work and API work has no per-surface limit in the pool itself, so that split has to be a convention.

This matters most for budgeting. A team that expects the Agent to do light work and the API to carry a nightly batch should size the plan to the total, not each part, and look at usage after the first week to see which one drove the spend. The API's usage endpoint shows what API jobs spent; the dashboard billing page shows the pool as a whole. Plan limits such as concurrency and queue depth are separate from the credit pool, so running out of slots is a different problem from running out of credit.