Search HTTP status codes by number or by what you are trying to do, with when to send each one and the misunderstanding that comes with it.
47 status codes
Search by number, by reason phrase, or by what you are trying to do.
The headers were received; send the body.
1xx · Informational
When to send it
Sent automatically when a client sends `Expect: 100-continue` before a large body, so it can learn the request will be rejected before uploading gigabytes.
What people get wrong
curl sends `Expect: 100-continue` on bodies over 1 KB by default, which adds a round trip against servers that do not answer it. That is the usual cause of a mysterious one-second pause on uploads.
Read next to
417 (4xx)
About the 1xx class
The request was received and the process is continuing. These are interim responses — a final response still follows, and most client libraries handle them without telling you.
HTTP Status Code Reference runs entirely on your device — nothing is uploaded.
You’re on the Free plan. Compare plans
A list of codes and reason phrases exists on a hundred sites already. What is harder to find in one place is the consequence of each, which is what this reference leads with.
The redirect codes are the clearest case. 301 is implemented by essentially every client as changing a POST to a GET, which means a redirected form submission silently becomes a read. 308 exists precisely to preserve the method, and 307 does the same for a temporary move. Picking between them by their names rather than their behaviour is how a payment endpoint quietly stops receiving data.
The 4xx pair people swap constantly is 401 and 403. 401 means unauthenticated — the name is simply wrong — so the caller should authenticate and retry. 403 means the server knows who you are and the answer is still no, so retrying with the same credentials is pointless. Returning the wrong one sends a client into a retry loop or out of one too early.
The other is 400 against 422. If the request would not parse, 400 is right. If it parsed and the values were unacceptable, 422 says so precisely, and most APIs returning 400 for validation should be returning 422 with a body naming the fields.
Search covers the prose as well as the number, so "redirect post" finds the three codes that question is really about.
Both are permanent. The difference is the method: historically clients change a POST to a GET on a 301, and 308 exists to preserve the method and body. If a redirected request must stay a POST, 308 is the one. Both are cached hard, so test with a temporary code first.
422 if the request parsed and the values were wrong — a malformed email, an end date before a start date. 400 if the server could not parse it at all. Most APIs use 400 for both, which loses the distinction a client needs to tell a retry from a fix.
401 means unauthenticated, despite being called Unauthorized: the client should supply credentials and try again, and the response must include WWW-Authenticate. 403 means authenticated and still not permitted, so retrying changes nothing.
Almost never over HTTP. Every proxy, cache, retry policy and monitor between you and the caller reads the status line rather than your JSON, so an error behind a 200 is invisible to all of them.
That a proxy stopped waiting — not that the work stopped. The origin may well have completed it. That is why retrying a non-idempotent request after a 504 can perform it twice.