Deployment rate limits
Each deployment has rate limits to protect your assistant from abuse and control cost. Limits apply per deployment — widget and Direct API deployments are counted separately.
Limit types
Section titled “Limit types”| Limit | Description |
|---|---|
| RPM | Requests per minute per deployment (messages, config fetches) |
| IP burst | Short-window cap per visitor IP on widget endpoints |
| Plan ceiling | Organization-wide caps from your billing plan |
Defaults and overrides
Section titled “Defaults and overrides”Default limits are set by plan tier. Enterprise customers may request higher RPM or custom burst rules — contact support.
When limits trigger
Section titled “When limits trigger”| Response | Visitor experience |
|---|---|
| 429 Too Many Requests | Widget shows friendly retry message |
| Retry-After header | Client should backoff |
See Rate limit troubleshooting.
Best practices
Section titled “Best practices”| Scenario | Recommendation |
|---|---|
| Marketing campaign spike | Raise limits or cache FAQ-heavy paths |
| Staging load tests | Use staging deployment with separate limits |
| Backend integration | Exponential backoff on 429 in your Direct API client |
Relation to tool calls
Section titled “Relation to tool calls”Message rate limits are separate from your upstream API quotas. Heavy tool usage still respects deployment RPM — monitor usage and tool logs.