Rate Limits

Control how fast users and agents can send requests. Applied per-user via Redis sliding window.

Global Limits

Requests per minute (per user)60

10300

Tokens per minute (per user)100,000

10K500K

• Rate limits are enforced in the orchestrator chokepoint before every agent turn.

• Tracked in Redis with a 60-second sliding window key: rl:{user_id}:{YYYYMMDDHHMM}

• Exceeding the limit returns HTTP 403 with code RATE_LIMIT.

• Kill switch bedrock_global overrides all limits — disabling it stops all traffic regardless of rate.