curl to Python

What this tool is used for

• Converting an API's curl example into requests code for a script.
• Reproducing a captured request inside an existing Python client.
• Getting the auth and header setup right before wrapping it in a function.
• Producing a snippet for a notebook or a one-off task.
• Checking how a form body differs from a JSON body in requests.

Frequently Asked Questions

Should I use requests or httpx?

requests is synchronous only and the de-facto standard. httpx has a near-identical API but adds async (await client.get(...)) and HTTP/2. If you are porting a curl command into an existing sync script, requests is the smaller change; if you need concurrency, httpx avoids bolting on threads.

Why does my request hang forever?

requests has NO default timeout — omitting the timeout argument means it can block indefinitely, which curl never does. Always pass timeout=(connect, read), e.g. timeout=(3.05, 27). This is the single most common production bug when porting from curl.

What is the difference between data= and json=?

data= sends form-encoded key/value pairs (application/x-www-form-urlencoded), matching curl's plain -d. json= serialises the object and sets Content-Type: application/json automatically. If your curl used -H 'Content-Type: application/json' -d '{...}', you want json=.

How does curl's -k / --insecure translate?

verify=False, which disables certificate verification and emits an InsecureRequestWarning. Prefer pointing verify at a CA bundle path instead: verify='/path/to/ca.pem'. Disabling verification in production defeats TLS entirely.

How do I reuse a connection across requests?

Use a Session: s = requests.Session() then s.get(...). This keeps the underlying TCP/TLS connection alive via urllib3 connection pooling and persists cookies between calls — the equivalent of curl's default behaviour within a single invocation.

Should the generated code use a Session?

For more than one request to the same host, yes — a `requests.Session` reuses the TCP connection and carries cookies, which is most of the latency saving. For a single call the plain function is equivalent.

Common errors and gotchas

• Using `data=` for a JSON payload, where `json=` sets the content-type and encodes it for you.
• Copying `-k` into `verify=False`, which disables TLS verification and emits a warning for good reason.
• Making many calls without a `Session`, which discards connection pooling entirely.
• Omitting a timeout, which by default means requests can hang indefinitely.
• Assuming a non-2xx raises, when you must call `raise_for_status()` yourself.