#is there a way to set a rate limit to API calls?

1 messages · Page 1 of 1 (latest)

dusky goblet
#

I'm using Mistral Large latest for testing and getting everything working.
Currently chatting through the TUI but since i'm using the free mistral option i have a rate limit of 1 per second for llm api calls which is resulting in errors, is there a way to set a 1 rps rate limit inside openclaw?

topaz helmBOT
#

To help others find answers, you can mark your question as solved via Right click solution message -> Apps -> ✅ Mark Solution

thin roseBOT
#

OpenClaw doesn’t currently have a “hard 1 request/sec” throttle for LLM API calls (esp. not inside a single agent run / tool loop). If the model does tool-use, one user message can easily cause 2+ back-to-back LLM calls (tool request → tool result → follow-up), which will trip a 1 RPS free-tier limit even if you only send one message.

What you can do today:

  1. Reduce burstiness between turns (helps if the errors are from you sending messages quickly)
  • In the TUI/session:
    • /queue collect debounce:2s
    • (or /queue followup debounce:2s)
  • Or in config:
    {
  messages: { queue: { mode: "collect", debounceMs: 2000 } }
}

Docs: https://docs.openclaw.ai/concepts/queue

  1. If you truly must stay on 1 RPS: you’ll need one of these workarounds