Rate limits

To keep the API fast for everyone, it limits how often it can be called. Two limits apply, counted separately:

600 requests a minute per API key.
600 requests a minute per IP address. Several keys used from one server share that server’s limit.

Each response includes a RateLimit-Limit header (the limit) and a RateLimit-Remaining header (requests left in the current one-minute window). Server code can read them; code running in a browser, including the playground, can’t, because the API doesn’t expose them to web pages. Over a limit, the API answers 429 with "code": "rate_limited" and a Retry-After header: the number of seconds to wait before trying again.

if (response.status === 429) {
  const seconds = Number(response.headers.get("Retry-After") ?? "1");
  await new Promise(resolve => setTimeout(resolve, seconds * 1000));
  // then send the same request again
}

To stay well under the limit, spread regular syncs out instead of sending bursts, and keep results that rarely change, such as a campaign’s settings, instead of fetching them every time. AI assistants connected through the MCP server have their own, separate limit. See MCP tools and limits.

Get started

AI assistants (MCP)

Using the API