# Rate limits

- Fair-use limit: **600 requests per minute per client IP** across `/api`, `/mcp` and `/a2a`, with short bursts tolerated. Discovery files (`/llms.txt`, `/openapi.json`, `/.well-known/*`, Markdown pages) are cacheable and not counted strictly.
- Headers on API responses: `X-RateLimit-Limit`, `X-RateLimit-Remaining` (approximate), `X-RateLimit-Reset` (Unix seconds).
- When exceeded: HTTP 429 with `Retry-After` (seconds) and a JSON error `RATE_LIMITED` (`retryable: true`, `retry_after_seconds`).
- Results are deterministic: cache them client-side (`cache.cacheable`), and batch related calculations by following `next_actions` instead of polling.
- Heavy or commercial use: the service is free; contact via https://tttkmbb.com/docs if you need more than the fair-use limit.