Rate limits
A rolling account-level limit shared across API keys, MCP, and the dashboard Playground.
Limits
Each account is limited to 60 authenticated requests per rolling minute, shared across its API keys, hosted MCP, and dashboard Playground. If you hit the limit, calls return 429; back off and retry.
Latency
AI-model calls depend on upstream model APIs (typically a few seconds). Identical queries hit a cross-customer cache and return in under 200ms.