Rate limits

Rate limit headers

Every /api/*response carries the standard trio. Successful responses report the platform-wide limiter's window; a 429 from a specific endpoint reports that endpoint's window.

HTTP/1.1 429 Too Many Requests
x-ratelimit-limit: 120
x-ratelimit-remaining: 0
x-ratelimit-reset: 1783520640
retry-after: 42

{ "ok": false, "error": "RATE_LIMITED", "message": "Too many requests. Please slow down.", "retryAfter": 42 }

Platform-wide limit

Before any route-specific limit, every API request passes a global edge limiter: 120 requests per minute per IP across all /api/* routes (sliding window, enforced in Redis across all instances). Self-hosted deployments can tune this with OMEGA_GLOBAL_RATE_LIMIT.

Per-endpoint limits

Core endpoints add their own windows on top of the platform limit. All windows are 60 seconds, sliding. API-key limits apply per key; IP limits apply per source address.

Other endpoints carry route-specific limits sized to their cost; the headers above always tell you the window you are in. If you need more sustained throughput than a published limit, contact us — raises are handled per deployment, not by silently lifting the shared limits.

Gateway limits by plan

The LLM gateway is the one surface with a per-plan rate split. The per-key limit depends on your organization's plan:

Rate limits are about instantaneous throughput. Monthly decision quotas and overage pricing are a separate, plan-based entitlement — see Pricing & Limits.

Handling 429s

async function callWithBackoff(fn: () => Promise<Response>, maxRetries = 3) {
  for (let attempt = 0; attempt <= maxRetries; attempt++) {
    const res = await fn();
    if (res.status !== 429) return res;
    const retryAfter = Number(res.headers.get("retry-after") ?? "1");
    await new Promise((r) => setTimeout(r, retryAfter * 1000 * (attempt + 1)));
  }
  throw new Error("Rate limited after retries");
}

Respect Retry-After rather than retrying immediately — the window is sliding, so hammering a 429 keeps you rate-limited longer.