Debug Grok API errors

A failed request returns a status code plus an error message. Check https://status.x.ai (RSS: https://status.x.ai/feed.xml) when the failure looks like an outage. Typical per-endpoint codes live in the API Reference.

4XX

Status	Endpoints	Cause	Solution
400 Bad Request	All		Check the request body or URL.
401 Unauthorized	All	Missing or invalid authorization token.	Send Authorization: Bearer <key>. Mint a key in the xAI Console.
403 Forbidden	All		Ask the team admin for permission.
404 Not Found	All		Check the body and endpoint URL against the API Reference.
405 Method Not Allowed	All	Wrong HTTP method (for example POST to a GET-only route).	Match the method in the API Reference.
415 Unsupported Media Type	POST endpoints
422 Unprocessable Entity	POST endpoints	A field in the POST body has an invalid format.	Check the body against the API Reference.
429 Too Many Requests	Inference	Request rate hit the limit.	Slow down, or raise the limit in the xAI Console.

2XX that still means wait

Status	Endpoints	Cause	Solution
202 Accepted	/v1/chat/deferred-completion/{request_id}	Deferred chat completion is queued; the response is not ready.	Poll until processing finishes.

Most error bodies explain themselves. For 429, pair this with send-async-concurrent-requests (semaphore) or the Batch API. For 202, see run-deferred-chat-completion.

Bug report

Email the API request, response, and relevant logs to support@x.ai. You can also ask in the #help channel of the xAI API Developer Discord.

Pitfalls

• 202 on the deferred-completion GET is success-but-not-ready, not a failure.
• 401 is almost always a missing Authorization: Bearer header or a bad key.
• 429 is a rate limit. The async docs say you cannot exceed the console cap with concurrency.
• Service status is also shown in the docs site navigation bar.
• Console API credits are separate from SuperGrok's weekly pool on grok.com.