Support --cookie-file in httpx[cli]

[cryptrr]

This is very important for automation use cases.

So imagine I want to send a POST request to an authenticated endpoint. I need to include a certain cookie with an ACCESS_TOKEN

You can do it manually but that ACCESS_TOKEN will expire and need to be updated again after fetching it again from the browser.

With a netscape style cookie file passed in as --cookie-file , the httpx library can update the new cookies received from every request and the cookie file acts as a cookie jar.

This feature works great in yt-dlp and the login cookies passed in never expires because it is being updated by yt-dlp.

[wellkilo]

HTTPX's Python API can persist cookies, but the command-line client does not currently expose a Netscape cookie-jar/file option.

For automation, use a small Python wrapper with a persistent cookie jar. HTTPX's Cookies object is compatible with Python's cookiejar model for requests, while MozillaCookieJar can load/save Netscape-format files:

import http.cookiejar
import httpx

jar = http.cookiejar.MozillaCookieJar('cookies.txt')
try:
    jar.load(ignore_discard=True, ignore_expires=True)
except FileNotFoundError:
    pass

with httpx.Client(cookies=jar, follow_redirects=True) as client:
    response = client.post('https://example.com/api', data={...})
    response.raise_for_status()

jar.save(ignore_discard=True, ignore_expires=True)

Be careful with assumptions about “never expires”: a server can invalidate/rotate an access token independently of cookie expiry, and HttpOnly/SameSite/domain/path rules still apply. Browser cookie databases are not Netscape files and may be encrypted; exporting them is a separate, security-sensitive step.

A future --cookie-file CLI flag should define load/save timing, session-cookie handling, file locking, permissions, and whether it overwrites the file after redirects. Until then, the Python API is the appropriate place for stateful authenticated automation; the httpx CLI is intentionally a lightweight request tool.