Usage, limits and errors

Understand the shared REST and MCP allowance and handle retries.

Shared organisation allowance

All API keys share the organisation's allowance. REST data requests and MCP tool calls count together. Scale includes 100,000 requests per calendar month and up to 120 requests per minute. Monthly counters reset on the first day of the month at 00:00 UTC. Minute limits use fixed minute windows. Additional network abuse protection can also return a rate limit.

Each accepted REST data request or validated MCP tool call uses one request. A 404 or upstream failure after acceptance still counts. Invalid credentials, missing scopes, invalid arguments and rejected quota requests do not count. MCP connection, initialization and tool discovery are not charged. No automatic overage charges are applied.

Response headers

Successful REST responses include RateLimit-Limit, RateLimit-Remaining and RateLimit-Reset (seconds until the minute window resets), plus X-Monthly-Limit and X-Monthly-Remaining. Settings → Developers shows this month's REST requests and MCP calls, as well as each key's lifetime request count and last-used date.

Retry limits

A 429 response includes Retry-After in seconds. If the code is monthly_limit_exceeded, it points to the next monthly reset. Otherwise wait for the shorter rate window. Use backoff and jitter for temporary failures; avoid retrying invalid requests or credentials.

{
  "error": {
    "code": "rate_limited",
    "message": "Your organisation's request rate limit has been reached.",
    "requestId": "request identifier"
  }
}

Include requestId when reporting a REST error. MCP tool errors return isError with a JSON text payload containing code, message and retryAfter when applicable.

Other limits

An organisation can have 20 active API keys. Search pages contain up to 100 results. The maximum search offset is 10,000 and search text is limited to 200 characters. MCP request bodies are limited to 16 KiB.