Rate limits
Limits are enforced at the CloudFront edge by API key (or agent token). Limits are per-minute and per-day; whichever fires first applies.
Every response carries X-RateLimit-Limit, X-RateLimit-Remaining and X-RateLimit-Reset (epoch seconds). On a 429, honour Retry-After in seconds.
Backoff guidance
- Reuse one MCP session (
Mcp-Session-Id) across calls instead of re-initializing per request. - Coalesce identical
resolve_locality/read_witnessescalls inside your service. - Read a cell once and reason over the full row, rather than issuing many narrow calls for the same subject.