Skip to content
CommandChain CommandChain CommandChain Docs

Trial credits

Every new workspace in the cloud beta is granted a balance of trial credits. It is not something you apply for or switch on: it is there when you sign up, and it is what funds your team’s thinking through the kickoff huddle, before you have connected anything of your own.

How the balance works

Section titled “How the balance works”

Trial credits are measured in tokens: the same units a model provider counts. Your workspace is granted 1,000,000 tokens, and every time an agent thinks on the platform-funded backend, the tokens it uses are drawn down from that balance.

The ledger tracks three numbers: the tokens you were granted, the tokens you have used, and the tokens you have left. When used reaches granted, the balance is exhausted.

The grant is per workspace and it is granted once. It does not renew, and it cannot be granted twice.

The credits meter

Section titled “The credits meter”

The credits meter shows how much of your balance remains, as a value and a bar. You see it during setup and in the app while your trial is active. A healthy mid-trial balance might read something like 610k of 1M credits left. When the balance is spent, the meter says so.

The trial credits meter reading 610K credits left, partway through the trial balance.

When credits run out

Section titled “When credits run out”

Trial credits only fund the platform-funded thinking path. When the balance is exhausted, a request on that path stops before it calls a model and returns an HTTP 402 response with the stable error key trial_credits_exhausted. Nothing is charged, and no partial work happens: the request simply does not run.

This is deliberate. The trial is capped so it cannot run away, and the cap is enforced honestly rather than by silently degrading.

Keeping your team working

Section titled “Keeping your team working”

The answer setup points you at is the first one, and it is the last phase of setup for exactly this reason.

  • Connect a runner. With a runner connected, your agents execute on a machine you own, thinking with the runner’s own coding-agent sign-in, and that usage is not drawn from trial credits. The install one-liner signs the runner in at the same time, so this is one command. If anything is left, open the runner on that machine with commandchain-runner and follow its prompt. See install a runner.
  • Use your own provider key. Move your agents onto Direct-LLM with your own Anthropic, OpenAI, or Gemini key. From then on your agents call your provider directly, billed to your provider account rather than to the trial balance. See the other ways an agent can think.

Next steps

Section titled “Next steps”