AI provider error reference

The errors AI coding tools actually return, what each one means, and how to get moving again. Each page also documents exactly what kRouter does on your behalf when it sees that error — how long the account is parked, whether it fails over, and when it comes back.

• Google Antigravity, Gemini Cloud Code Assist

  Antigravity "Verify your account to continue" — what it means and how to fix it

  Google's anti-abuse system has flagged the account and put a human-verification gate in front of it. It is not a quota problem and it will not clear on its own — every request on that account returns 403 until you complete the check Google is asking for.

  Read the fix

• Google Antigravity, Gemini, and other Google-backed providers

  AI provider "account has been suspended" or "service has been disabled" — recovery steps

  The provider has taken the account out of service. Unlike a rate limit this is a policy decision, so retrying — from any tool, not just kRouter — will keep failing until the account itself is restored.

  Read the fix

• Kiro (AWS Q Developer), Amazon CodeWhisperer

  Kiro "You have reached the limit" / MONTHLY_REQUEST_COUNT — what to do

  The account has spent its monthly request allowance. This is a billing-cycle limit, not a per-minute rate limit, so waiting a few minutes does nothing — it resets when the next cycle starts.

  Read the fix

• Kiro (AWS Q Developer)

  Kiro CONTENT_LENGTH_EXCEEDS_THRESHOLD — "Input content length exceeds threshold"

  The request body itself is larger than the provider will accept. The problem is the request, not the account — which matters, because it means switching accounts cannot help.

  Read the fix

• OpenAI, Anthropic, Google, and most OpenAI-compatible providers

  "Context length exceeded" / "prompt is too long" — fixing oversized AI requests

  The prompt plus the requested output is larger than the model's context window. The model never starts generating — the request is rejected up front.

  Read the fix

• Every provider — OpenAI, Anthropic, Google, Groq, and the rest

  "Rate limit exceeded" / 429 Too Many Requests — how kRouter keeps the request alive

  You have sent more requests, or more tokens, than the provider allows in the current window. It is temporary: the window rolls and the account recovers on its own.

  Read the fix

• All providers

  "Model not found" / "Requested entity was not found" — fixing a bad model id

  The provider does not recognise the model id you sent. Nothing is wrong with your account or your quota — the name is wrong, retired, or not available on your plan.

  Read the fix

• Anthropic, Google, and other capacity-constrained providers

  "No capacity available" / "Overloaded" — provider is out of compute

  The provider has no free capacity for that model right now. It is their infrastructure, not your account — an identical request from anyone else would fail the same way.

  Read the fix

• All providers

  "Request too large" / "Payload too large" (413) — trimming an oversized request

  The HTTP body exceeded the provider's size cap. This is a transport limit measured in bytes, which is not the same thing as the model's token context window — you can trip it with a large attachment even on a short conversation.

  Read the fix

• All providers

  "No credentials" / "No active credentials for provider" — connecting an account

  kRouter has no usable account for the provider you addressed. Either none is connected, or every connected one is currently locked, banned or exhausted.

  Read the fix