Hermes disables a valid key on "overloaded" responses
Hermes treats provider 'overloaded' responses as rate_limit with should_rotate_credential=True, so a perfectly valid API key gets marked exhausted after just two errors and every retry then fails. Add a second key so the credential pool survives a transient overload, or lengthen the cooldown. The proper fix - a dedicated overloaded class - is upstream.
Likely cause
Hermes classifies provider 'overloaded' responses as rate_limit with should_rotate_credential=True, so a single valid API key is marked exhausted after just two errors and every retry then fails.
The fix
- 1 Add a second API key for the provider so the credential pool survives a transient overload.
- 2 Lengthen the cooldown so a key isn't retired after only two errors.
- 3 The proper fix (a dedicated overloaded failover class) is upstream - track the issue.
Hit a different error?
Paste any agent error and get the cause and fix in seconds.
Frequently asked questions
Why does one valid key get marked exhausted so fast?
Hermes counts 'overloaded' responses as rate-limit failures and rotates the credential after two of them. With a single key there's nothing to rotate to, so all retries fail. Add a second key so the pool survives.
I only have one API key. What can I do?
Lengthen the retry cooldown so a transient overload doesn't burn through the two-error threshold, and retry the request a bit later. A second key on the pool is the more robust workaround.
Is this the same as a real rate limit?
No. 'Overloaded' is a transient server-side capacity signal, not your quota. Hermes conflating the two is the bug; the upstream fix adds a separate overloaded class so valid keys aren't retired.
Stop firefighting agent errors
Decoding errors one at a time is the manual version of what BetterClaw automates. Run your OpenClaw agents hosted with managed models, retries and config validation built in.
$19/month per agent · BYOK · 7-day money-back guarantee