Get Started

Rate Limits

If you require higher limits, please contact us at support@caesar.xyz.

Scope: These limits apply to research job creation endpoints:

  • POST /research (native API)
  • POST /compat/completions (OpenAI-compatible)
  • POST /compat/chat/completions (OpenAI-compatible)

The Caesar API has the following limits:

  • Concurrent research jobs: 5 per account (in-progress jobs, including queued).
  • Monthly reasoning budget: 200 reasoning loops, counted only for successful research jobs.

Usage is based on actual work performed. reasoning_loops_consumed reflects the actual loops used, not reasoning_loops requested. If you set reasoning_loops: 5 but the job completes in 2 loops (via early exit or sufficient information), only 2 loops count toward your usage.

Exceeding limits

When a limit is exceeded the API returns 429 Too Many Requests:

1{ "error": "Too many requests - rate limit exceeded." }

Recommended client behaviour

  • Retry with exponential back-off.
  • Serialise new jobs when you’re at the concurrency cap.
  • Use reasoning_loops thoughtfully to stay within your monthly budget.

A reasoning loop represents one iteration of Caesar’s research process: gathering information, analyzing findings, and refining the response. Your monthly usage is the sum of reasoning_loops_consumed by jobs that reach the completed status, not the reasoning_loops requested.

Any job that is not yet completed or failed (e.g. queued, searching, summarising, analysing) counts towards your concurrency limit.

Limits reset on a monthly cadence; your dashboard shows your remaining balance and the next reset window.

Yes. Email support@caesar.xyz with your expected volumes and use case.

Visit your dashboard after signing in.