Rate limits and complete exports
Per-user limits for each endpoint and how to export a full date range.
Default limits
| Endpoint | Limit | Shared budget |
|---|---|---|
| Session list | 2 requests per 4 seconds | Shared by v5 and v6. Pace calls at least 2.5 seconds apart. |
| Metric details | 4 requests per 4 seconds | Shared by v6 and v7. |
| Session detail | 1 request per second | |
| Visible directory | 2 requests per 4 seconds | |
| Bootstrap | 2 requests per 4 seconds |
Complete exports
- Session-list returns the newest matching rows, sorted by started_at descending. There is no cursor or offset.
- Use a maximum limit of 1000. If a time window reaches the cap, divide it into smaller [start, end) windows and deduplicate by session_id. Dense windows may also require rep filters. Do not label a capped or sampled response as a complete export.
- Default limits apply per user; coordinate workers sharing a key. Account overrides and upstream protection may differ.
- On 429, wait the Retry-After delay in seconds. Use bounded backoff for transient 5xx responses and read errors as text.
A response containing 1,000 rows may have reached the cap. It does not establish that you received every matching call. Use aggregates for totals and clearly label capped or sampled lists.
What a 429 response means is on Errors and troubleshooting.