Every HTTP status code, grouped by class and searchable by number or by word. The descriptions are written for the reason people actually look a code up, which is rarely to learn its official name: usually you are choosing between two codes, or one has turned up and you need to know which end of the connection to investigate. So 401 says plainly that it is about authentication and 403 is the one for a user who is logged in but not allowed. 502 versus 504 is spelled out as answered-badly versus did-not-answer-in-time, which tells you whether to look at your application or at how slow it is. 422 explains when it beats 400. 301 warns that browsers may cache it for months. Type a single digit to browse a whole class. Rarely-used codes are tucked behind a toggle but still turn up in search.
| Situation | Use |
|---|---|
| Not authenticated | 401 |
| Authenticated but not allowed | 403 |
| Malformed request body | 400 |
| Valid syntax, invalid content | 422 |
| Upstream answered badly | 502 |
| Upstream too slow | 504 |
| Permanent move, method preserved | 308 |
| Redirect after a form POST | 303 |
401 means the request was not authenticated — no credentials, or invalid ones — and retrying with valid credentials may work. 403 means the server knows who you are and is refusing anyway; re-authenticating will not help. The name 'Unauthorized' on 401 is the source of the confusion, since it is really about authentication.
Both come from a proxy in front of your application. 502 means upstream answered with something invalid — usually the app crashed or is not listening. 504 means upstream did not answer in time — the app is running but too slow. That distinction tells you whether to look at errors or at performance.
400 if the request could not be understood at all — malformed JSON, bad framing. 422 if the syntax was fine and the content was wrong, such as a missing required field. 422 is the more precise answer once the body has successfully parsed.
308 for a permanent move, 307 for a temporary one — both preserve the method and body. Use 303 to redirect after a form POST so a refresh does not resubmit. Avoid 301 unless you are certain, because browsers cache it aggressively and it is hard to undo.
It is an April Fools' joke from 1998 that has outlived several attempts to remove it. Some services return it deliberately to mean 'refused, and not going to explain'.
Because they exist but are effectively unused outside WebDAV, and showing all of them makes the list harder to scan. Toggle them on, or just search for one directly.
This is reference data, so the engineering is in what the descriptions say rather than in any computation. Almost nobody looks up a status code to find out what the RFC calls it — they look it up because they are deciding between two codes, or because one has appeared and they do not know which side of the connection to investigate. The descriptions are written for that. Where a code is routinely confused with a neighbour the entry names the alternative: 401 explains that it is about authentication and 403 is the one for a permitted user who is not allowed; 403 notes that 404 is sometimes the better answer when revealing existence is itself a leak; 422 explains it is more precise than 400 once the body has parsed; 301 warns that browsers cache it for months and 302 or 307 is safer if reverting is possible; and 502 against 504 is spelled out as answered-badly versus did-not-answer-in-time, which is the distinction that tells you whether to look at the application or at its speed. Search accepts either a number or text, and a numeric query matches by prefix so typing 4 lists the whole client-error class — that is how people browse when they half-remember a code. Codes that exist but are effectively unused, mostly WebDAV, are hidden behind a toggle so the default view is the set worth knowing, while a direct search still finds them. Everything is static and local; nothing is fetched.