Error pages
LLM API error fixes by provider and code.
Each page explains what the error means, whether it is likely your code or provider-side, and the exact cURL test to run.
OpenAI API Error 429: Rate Limit or Quota Exceeded
OpenAI API 429 usually means your request is being rate limited, your quota is exhausted, or your project has hit a per-minute token/request cap.
rate_limit mixed
Anthropic API 529: Overloaded Error
Anthropic API 529 means the provider is overloaded or temporarily unable to serve your request at normal capacity.
provider_overload provider/server side
OpenAI API 401: Invalid API Key
OpenAI API 401 means the API key is missing, malformed, revoked, or not valid for the project you are calling from.
auth your key or account
OpenAI API 403: Forbidden or Model Access Denied
OpenAI API 403 means the key is recognized but the account, project, model, region, or policy does not allow the requested operation.
permission your key or account
Gemini API Fetch Failed: Network, CORS, or Endpoint Issue
Gemini API fetch failed usually means your runtime could not complete the HTTP request because of network, CORS, DNS, endpoint, or key placement problems.
client_runtime network path
Claude API ECONNRESET: Connection Reset During Request
Claude API ECONNRESET means the TCP connection was closed unexpectedly by a peer, proxy, runtime, or network path before the response completed.
network network path
OpenRouter 402: Insufficient Credits
OpenRouter 402 means the request cannot be billed because the account lacks credits, has a payment issue, or is blocked by spend controls.
billing your key or account