Rate limits

Rate limits protect your workspace and the platform. Once a request is authenticated, limits are counted per workspace: every API key of the same workspace shares the same budgets. Requests that are not yet tied to a workspace are counted per client IP.

All limits use a 60-second window. The counter starts with the first request in the window and resets when the window ends.

General API limit

Most /api/* routes are behind a general limit that scales with your plan:

Stricter limits on specific routes

Some routes carry a stricter limit. Where it is added on top of the general limit, a request must pass both. Calls, WhatsApp, email and billing routes are governed by their own limit instead of the general one.

Response headers

Responses from rate-limited routes include:

When several limits apply to one request, the RateLimit-* headers describe one of them. Always rely on Retry-After when you receive a 429.

When you hit a limit

The API answers 429 Too Many Requests with code RATE_LIMIT_EXCEEDED:

HTTP/1.1 429 Too Many Requests
Retry-After: 42
RateLimit-Limit: 120
RateLimit-Remaining: 0
RateLimit-Reset: 42
Content-Type: application/json

{
  "error": {
    "code": "RATE_LIMIT_EXCEEDED",
    "message": "Too many requests, please slow down"
  }
}

The message differs between limits (for example Rate limit exceeded for outbound operations or Too many agent operations, please slow down); the code is always RATE_LIMIT_EXCEEDED.

Handling 429 responses

async function jelliuFetch(url, init, attempts = 5) {
  for (let i = 0; i < attempts; i++) {
    const res = await fetch(url, init);
    if (res.status !== 429 && res.status !== 503) return res;
    const retryAfter = Number(res.headers.get('Retry-After'));
    const waitMs = Number.isFinite(retryAfter) && retryAfter > 0
      ? retryAfter * 1000
      : 2 ** i * 1000;
    await new Promise((resolve) => setTimeout(resolve, waitMs));
  }
  throw new Error('Rate limited: retries exhausted');
}

Other limits

• MCP server. The MCP endpoint allows 120 requests per minute per client IP. Each tool the assistant calls is an API request and also counts against your workspace limits above.
• Web chat widget. Public widget traffic is limited per visitor IP, per widget (its rate_limit_rpm setting, 30 by default) and per workspace (300 per minute across all widgets). See Web chat widget.
• Plan capacity such as concurrent calls, and budgets you configure on agents, are not rate limits. They return their own codes (MAX_CONCURRENT_CALLS_REACHED, AGENT_BUDGET_EXCEEDED); see Errors.