Starter
A clean cloud balance for the runs that need a separate machine. Local reviews remain included.
- Independent cloud finish runs
- Proof artifacts + history
- Local CLI stages included
Critique's pinned reviewer runtime is included locally. Cloud adds a 15¢ execution fee per stage when you need a separate clean machine; managed model usage is reported separately.
Cloud stages are shared across your account at fifteen cents each. Local execution does not consume this balance.
A clean cloud balance for the runs that need a separate machine. Local reviews remain included.
For agents that finish work continuously across local branches, cloud sandboxes, and CI.
A shared meter for teams making Critique part of their shipping contract.
Run locally through Critique's pinned reviewer runtime, use managed Critique inference, or choose a clean cloud machine. Critique never hides model spend inside the check price.
Cloud stage packs cover Critique orchestration. Local stages are included. Model usage is always shown and billed independently by the selected runtime or provider.
A receipt separates execution location, Critique’s cloud fee, input/output tokens, and model cost. A costly model never disappears behind a cheap-looking stage price.
Read the finish contract →Use the runtime you already trust. Local execution is included; cloud orchestration and model usage remain visible as separate lines.
One cloud check is one Critique execution stage: the remote isolated workspace, change capsule, independent agent contract, and proof artifact. Cloud stages cost $0.15. Local Critique reviewer stages are included.
No. The 15¢ is the Critique Cloud execution fee. Managed cloud model usage is a separate charge. Local Critique stages have no execution fee; when you are signed in, model tokens bill your Critique credit balance via the Inference API. Otherwise the model provider you configure bills usage directly.
Critique starts with the task and the exact work in progress, then follows nearby code and tests when they matter. Local review stays in a separate workspace on your machine. When you explicitly choose cloud, Critique uploads the protected current working tree—including uncommitted changes—and gives each critic, repairer, and verifier a fresh full-repository workspace.
Coding agents change the unit of work. A team with three developers can generate more candidate changes than a team of thirty. Critique prices the independent finish pass, not the number of humans who happen to be on the account.
Critique can run a critic, a bounded repair, and a fresh independent verifier. Local stages are included. Each cloud stage is 15¢, and managed model usage is billed separately. The receipt records execution, tokens, and model cost independently.