cURL Converter

Convert curl -u user:pass -d 'name=Ada' https://api.example.com/users to Python requests.

Tokens: the shell splits it into 6 words: curl, -u, user:pass, -d, name=Ada (the quotes are removed) and the URL.
Method: there is no -X, so the -d decides it: POST.
Headers: -d also implies Content-Type: application/x-www-form-urlencoded, and -u becomes Authorization: Basic + base64 of the 9 bytes user:pass = dXNlcjpwYXNz.
Code: requests.post('https://api.example.com/users', headers=headers, data=data, auth=('user', 'pass')), with the Content-Type in headers and data = 'name=Ada' sent exactly as written.

Add -G and the data moves into the URL instead: curl -G --data-urlencode 'q=hello world' https://api.example.com/search is a GET to https://api.example.com/search?q=hello%20world.

The common mistake: sending JSON with plain -d. curl -d '{"name":"Ada"}' https://api.example.com/users sends the 14 bytes of JSON labelled application/x-www-form-urlencoded, so many APIs answer 415 Unsupported Media Type or read an empty body. A converter that silently writes json= in Python hides the bug in curl and changes the request. This one keeps the request as curl sends it and warns you; add -H 'Content-Type: application/json', or use --json (curl 7.82.0 and later), which also adds Accept: application/json.

curl grew out of httpget, a small program by Rafael Sagula that Daniel Stenberg started improving in 1996. It was renamed urlget, and on 20 March 1998 it was released as curl 4.0, the name meaning “client for URLs”. Stenberg still leads the project. The library behind the tool, libcurl, followed in 2000 and is what PHP’s cURL extension wraps, which is why the PHP output here uses the same option names as curl itself.

Because curl ships with Linux distributions, macOS and, since 2018, Windows 10, API documentation adopted it as the neutral way to show a request: one line that runs in any terminal, with no language to choose. Browser developer tools followed with “Copy as cURL”, which is why most real commands arrive with a dozen headers and bash’s $'…' quoting.

The libraries it is converted to come from different eras: Python’s requests (2011) made auth= and json= one-word options, the WHATWG Fetch standard gave browsers a promise-based successor to XMLHttpRequest, and Node.js 18 (2022) made the same fetch() available on the server. Their defaults differ from curl’s in small ways, such as following redirects, and those differences are what the notes under the code point out.

This converter reads a curl command the way a shell does — quotes, escapes, $'…' strings, line continuations and Windows ^ escapes — then applies curl’s own rules: which flag sets the method, which headers are implied, how -d pieces are joined and what -G moves into the URL. It writes the request for seven targets and keeps bodies byte for byte, writing JSON as a literal only when no number, key order or duplicate key would change. In the other direction it reads a fetch() call, including Chrome’s “Copy as fetch”, with a parser that accepts literal values only.

Nothing is run and nothing is sent: commands often contain API keys, cookies and tokens, and they stay in your browser. Input up to 1 MB is converted as you type.

Related tools: JSON Formatter, HTTP Status Codes, and Base64 Encoder.