Usage and cost caps
Each project has caps on Qodex-funded LLM spend. The caps protect your project from accidental runaway usage. They apply only when Qodex is paying for the provider calls. Calls on the project’s own OpenAI key (BYOK) or its own connected ChatGPT subscription are paid by your account and never count against the Qodex-funded caps.How the caps work
The two rolling windows have no fixed reset. Headroom returns continuously as old usage leaves the window, so you do not need to wait for midnight or Monday. The countdown tells you when enough usage will roll off.
Other monthly quotas, such as included pull request reviews, also count per calendar month in UTC.
Where to see usage
Open the account menu at the bottom of the left navigation, then choose Usage. You will see:- The current plan, next to the page heading.
- A 5-hour limit bar.
- A Weekly bar.
- A Monthly bar.
What happens at the cap
When a project reaches its cap:- New chat runs pause with a clear cap message.
- Scheduled runs and PR reviews wait until headroom returns.
- In-progress runs stop at a safe boundary, save partial work, and mark the run as
budget_exceeded.
budget_exceeded is separate from a test failure. It means the run stopped because of budget, not because the app failed.
Per-run budget
Each test run also has its own token budget. The project cap protects the whole project. The per-run budget protects one scan from consuming too much context or cost by itself. At 80%, Qodex warns. At 100%, the run stops, saves partial output, and marks the runbudget_exceeded.
How to get unstuck
If you hit a cap:- Wait for usage to roll out of the window, or for the new month.
- Add BYOK so new calls route through your provider key.
- Connect your own ChatGPT subscription in Settings > LLM keys.
- Upgrade the project plan.
Next steps
Plans and pricing
See each plan’s cap.
BYOK
Use your own provider key.
Why is my scan slow?
Separate cost caps from scan runtime.
LLM cost on reruns
Learn why reruns usually cost less.