Rate limits and caps
On this page
Per key and per agent #
rate_limited with Retry-After in seconds and a message such as “60 requests a minute per key. Try again in 12 s.”
The limit is per credential, not per workspace. There is no workspace-wide request limit. An MCP tool call is one request to /mcp/v1. The REST routes a tool calls inside the server are not counted again.More than 20 refused tokens a minute from one client address are answered 429 whatever token is sent.
curl --retry 4 --retry-all-errors "$BRENZURI_URL/api/v1/sites" \
-H "Authorization: Bearer $BRENZURI_KEY"Daily credit caps #
dailyCap, 0 to 1000 credits. The day is UTC. What counts is what the credential held since 00:00 UTC minus what it released. Over the cap the answer is 429 budget_exhausted, with resetsAt, askAHuman and a Retry-After for the seconds until the reset. The first time a credential reaches its cap in a day the agent.budget_exhausted webhook is raised once.
Shared with people #
| Allowance | Limit | Refusal |
|---|---|---|
POST /sites/{siteID}/analyze) | rate_limited | |
POST /sites/{siteID}/topics/check) | rate_limited | |
rate_limited | ||
plan_limit |
OAuth endpoints #
| Endpoint | Per address | Per network |
|---|---|---|