#error Embedded agent failed before reply

1 messages · Page 1 of 1 (latest)

misty edge
#

All of sudden i am getting error:
05:12:18 error Embedded agent failed before reply: All models failed (1): anthropic/claude-opus-4-5: Provider anthropic is in cooldown (all profiles unavailable) (rate_limit)

carmine otterBOT
#

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

fair juncoBOT
# misty edge All of sudden i am getting error: 05:12:18 error Embedded agent failed before re...

That error means OpenClaw hit an Anthropic rate limit, so it put all your Anthropic auth profiles into a temporary cooldown (circuit-breaker behavior). Since your only configured model is anthropic/claude-opus-4-5 (and you likely have no fallbacks), the embedded agent has nowhere else to go → “All models failed”.

What to do:

  1. Wait a bit
    Cooldown backoff is typically 1 min → 5 min → 25 min → 1 hour depending on repeated failures. It should recover automatically once the cooldown expires.

  2. Verify cooldown + see when it clears

openclaw models status --json

Look for auth.unusableProfiles and the cooldownUntil timestamp.

  1. Prevent downtime next time (recommended): add fallbacks
  • Pick a fallback from your configured models:
# 
  openclaw models list
  • Then add one or more fallbacks:
    openclaw models fallbacks add <modelId>

(You can also set this in config via agents.defaults.model.fallbacks.)

  1. If it happens constantly, you’re probably over quota / too much concurrency for your Anthropic plan/key. In that case, adding a second provider fallback is the most reliable fix.

Docs that describe this behavior:

If you paste the auth.unusableProfiles block from openclaw models status --json (and openclaw --version), I can tell you exactly how long the cooldown is and whether it looks like a transient 429 burst vs a hard quota limit.

misty edge
#

aded a fallback but looks like antropic get back onlin emeanwhile

fair juncoBOT
# misty edge aded a fallback but looks like antropic get back onlin emeanwhile

Nice — that’s exactly what you want to see.

  • Anthropic came back because the cooldown expired (rate limits are usually temporary).
  • Adding a fallback is the right long-term fix, because next time Anthropic cools down, OpenClaw can automatically switch to the next model instead of failing the whole run.

If you want to double-check it’s set correctly:

openclaw models fallbacks list
openclaw models status --plain

And if this solved it, please right-click the solution message → Apps → ✅ Mark Solution so Answer Overflow indexes it for the next person.

misty edge
#

Fallbacks (1):

  • moonshot/kimi-k2-0905-preview
fair juncoBOT
# misty edge Fallbacks (1): - moonshot/kimi-k2-0905-preview

Yep — that output means your fallback is configured correctly:

Fallbacks (1):

  • moonshot/kimi-k2-0905-preview

So your routing is now effectively:

  1. try anthropic/claude-opus-4-5
  2. if Anthropic profiles are in cooldown / rate-limited → fail over to moonshot/kimi-k2-0905-preview

Two small recommendations:

  • Consider adding a second fallback (ideally a different provider) so you’re covered if Moonshot is also having a bad day:
    openclaw models list
openclaw models fallbacks add <anotherModelId>