Request logs
Logs lists a project's recent requests, newest first, so you can check what was sent, what it cost, and what went wrong.
What each entry shows
| Field | Meaning |
|---|---|
| Model, path, method and status | Which model and endpoint were called, and the HTTP status returned. |
| Created | When the request arrived. |
| Duration | From receiving the request to the last byte of the reply. |
| Cost | What the request cost you, with any discount applied. In a Token Plan workspace this is Credit consumed, at list price. |
| Saved and Discount | What your discount took off the list price. |
| Input, output and total tokens | Prompt and reply size. |
| Cached tokens | Prompt tokens served from cache at the lower rate. |
| Reasoning tokens | Tokens a reasoning model spent thinking. |
What isn't logged
Logs record what your bill needs: the model, endpoint, status, token counts, timing and cost. Prompts and completions are not stored — see privacy and data.
Retention
Request logs are kept for 90 days. Analytics keeps its totals for longer, so trends past 90 days remain visible.
Debugging with logs
- A request failed — find it by time and model; the status tells you which kind of error it was.
- A bill looks high — sort out which model and key it came from, then compare input, cached and output tokens. A low cached-token count on a long agent session usually means the prompt prefix changes between requests.
- A request is missing — check you're looking at the project the key belongs to.