#Claude Sonnet Rate Limit

1 messages · Page 1 of 1 (latest)

vital pawnBOT
#

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

torpid sail
#

Here is the issue I have from the logs follow command

run error: ⚠️ API rate limit reached. Please try again later.
connected | error
agent main | session main (openclaw-tui) |
anthropic/claude-sonnet-4-20250514 | tokens 31k/200k (15%)

safe hareBOT
#

That message is almost always Anthropic returning a 429 rate_limit_error. It’s not “messages per minute”, it’s usually tokens/min (and sometimes requests/min) on your Anthropic credential — so you can hit it with very few messages if each turn is large.

Two big clues in your snippet:

  • tokens 31k/200k (15%) → your current session context is already ~31k tokens. One request with a 31k-token prompt can exceed many Anthropic TPM limits by itself.
  • OpenClaw can also do extra hidden calls (auto-compaction retry, tool calls, retry-on-error), which makes it feel like “I only sent one message”.

Quick things to try (fastest signal)

  1. Reset the session (drops the 31k context):
  • send /new (standalone message)

Docs: https://docs.openclaw.ai/concepts/session (see “Reset triggers: exact /new or /reset … start a fresh session id”)

  1. Or compact instead of nuking context:
  • send /compact (standalone message)
  1. Wait ~1–5 minutes and retry (if you truly hit a per-minute window).
#

Verify which 429 you’re hitting

Please paste:

openclaw --version
openclaw models status
openclaw status --all

(status --all is designed to be shareable/redacted.)

And in openclaw logs --follow, look for the actual Anthropic error text. If it says:

General FAQ entry on Anthropic 429s:
https://docs.openclaw.ai/help/faq#why-am-i-seeing-http-429-ratelimiterror-from-anthropic