Keep getting this error | Friends of the Crustacean 🦞🤝 | Page 1

queen valleyBOT Feb 14, 2026, 6:27 PM

#

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

kind peakBOT Feb 14, 2026, 6:28 PM

#

That error means OpenClaw tried Anthropic, but every Anthropic auth profile you have is currently in cooldown (usually from rate_limit, sometimes timeout/billing/auth failures). With no usable Anthropic profile left, both configured Anthropic models fail → “All models failed”.

Do this:

Check which profile(s) are cooling down + why

openclaw models status --json

Look for auth.unusableProfiles (it’ll include the reason + cooldown time).

Wait out the cooldown (OpenClaw uses exponential backoff), or add another Anthropic profile so it can rotate:

Docs: https://docs.openclaw.ai/providers/anthropic (see “No available auth profile (all in cooldown/unavailable)”)

Add a non-Anthropic fallback so you’re not hard-down when Anthropic rate-limits
Docs: https://docs.openclaw.ai/concepts/models + https://docs.openclaw.ai/help/faq (Model failover section)
If this is happening “immediately”, paste the last ~30 lines around the failure from:

#

openclaw logs --follow

Also: what’s your openclaw --version? Cooldown behavior + failover got tweaks across versions.

kind peakBOT Feb 15, 2026, 10:17 PM

#

That message isn’t a “crash” — it’s OpenClaw telling you it ran out of usable models:

Anthropic got an HTTP 429 / rate_limit, so OpenClaw put the Anthropic provider in cooldown (exponential backoff). When it says “all profiles unavailable” it means it tried every Anthropic auth profile you have and all are currently rate-limited/cooldown’d.
Then it tried your fallback (moonshot/kimi) and that also hit a rate limit.
With both models failing, you get: “Agent failed before reply: All models failed”.

What to do:

Wait a bit and retry (cooldowns usually clear after the provider window resets).
Confirm you actually have quota on both providers (Anthropic subscription/setup-token especially can hit tight limits; API keys can also be capped/billing-limited).
Add either:

More auth profiles for the same provider (so profile rotation can keep working), and/or
A third fallback model/provider that has quota, so “All models failed (2)” becomes “(3)” and you still get replies when one provider is throttled.

Debug info that will make this 10× faster (safe to paste; it redacts secrets):

#

openclaw --version
openclaw status --all
openclaw models status
# optional (shows cooldown timers / unusable profiles)
openclaw models status --json

Docs refs (these sections describe exactly what you’re seeing):

kind peakBOT Feb 16, 2026, 6:59 PM

#

You’re hitting two separate auth problems during failover:

Anthropic: HTTP 401 … Invalid bearer token
That means the Anthropic credential you have stored is just invalid (revoked/typo/wrong token type).

Fix (Anthropic):

If you’re using Claude subscription auth: generate a fresh setup-token:

claude setup-token

then on the gateway host paste/store it:

openclaw models auth paste-token --provider anthropic

(or run openclaw models auth setup-token --provider anthropic to be prompted)

If you’re using an API key instead: make sure ANTHROPIC_API_KEY is set on the gateway host (best: ~/.openclaw/.env if you run the gateway as a service), then restart the gateway.

#

Docs: https://docs.openclaw.ai/help/faq#how-does-anthropic-setuptoken-auth-work

Google Gemini fallback: No API key found for provider "google"
OpenClaw is trying Gemini because it’s in your fallbacks, but you haven’t configured Google auth for that agent.

Fix (Gemini):

Either set GEMINI_API_KEY (again: put it in ~/.openclaw/.env if running via launchd/systemd), or
Remove/avoid Google models in agents.defaults.model.fallbacks so failover doesn’t route to Gemini.

Docs: https://docs.openclaw.ai/concepts/model-providers (Google provider uses GEMINI_API_KEY)
FAQ section: https://docs.openclaw.ai/help/faq#model-failover-and-all-models-failed

One extra gotcha: auth is per-agent
Your error shows it’s looking in:
/Users/yura/.openclaw/agents/main/agent/auth-profiles.json
So make sure you’re configuring that same agent (common issue: you authenticated a different agent, so this one is empty/wrong).

#

Docs: https://docs.openclaw.ai/help/faq#auth-profiles-what-they-are-and-how-to-manage-them

If you paste the output of:

openclaw --version
openclaw models status --probe

I can tell you exactly which provider/profile is failing and why.

violet carbon Feb 16, 2026, 7:05 PM

#

try to set a fallback if rate limit hits, for example kimi

kind peakBOT Feb 16, 2026, 7:05 PM

#

Yep — Kimi (Moonshot) is a good “rate-limit escape hatch” fallback because it’s a different provider than Anthropic.

Option A (CLI, easiest): add Kimi as a fallback + remove Gemini

openclaw models fallbacks list

openclaw models fallbacks remove google/gemini-2.5-flash
openclaw models fallbacks add moonshot/kimi-k2.5

openclaw models fallbacks list
openclaw gateway restart

Make sure Moonshot auth is actually usable

You already show MOONSHOT_API_KEY is present, but your probe timed out once. Test just Moonshot:

#

openclaw models status --probe-provider moonshot

If that still times out, it’s usually network/DNS/firewall on the gateway host (or the provider endpoint is having a moment).

Important: fix the broken Anthropic profile too

Your probe shows anthropic:manual is 401 invalid, so Anthropic can fail even before fallback kicks in. Lock Anthropic to the good profile:

openclaw models auth order set --provider anthropic anthropic:default
openclaw gateway restart

Docs (failover + fallbacks): https://docs.openclaw.ai/help/faq#model-failover-and-all-models-failed

#

That models status --probe output is super useful — it basically explains the failure:

1) Anthropic is fine… except your `anthropic:manual` profile is broken

anthropic:default (oauth) → ok
anthropic:manual (token) → HTTP 401 Invalid bearer token

So whenever OpenClaw rotates/tries anthropic:manual, the run can fail (and it can also contribute to “provider in cooldown / all profiles unavailable” if it keeps retrying a bad profile).

Fix options (pick one):

A) Stop trying the bad profile (recommended):

openclaw models auth order set --provider anthropic anthropic:default

(Optional: check current ordering first)

#

openclaw models auth order get --provider anthropic

B) Replace/fix anthropic:manual with a real token/key:

openclaw models auth paste-token --provider anthropic --profile-id anthropic:manual

Then paste a fresh Anthropic API key or a fresh claude setup-token.

2) Your fallback includes Gemini, but Google auth is missing

You have:

Fallbacks: ... , google/gemini-2.5-flash
Missing auth: google (GEMINI_API_KEY not set)

#Keep getting this error

Option A (CLI, easiest): add Kimi as a fallback + remove Gemini

Make sure Moonshot auth is actually usable

Important: fix the broken Anthropic profile too

1) Anthropic is fine… except your anthropic:manual profile is broken

2) Your fallback includes Gemini, but Google auth is missing

1) Anthropic is fine… except your `anthropic:manual` profile is broken