Sessions

Capture a login

1. Open Settings → Sessions. Enter the login URL — for example https://github.com/login.
2. Press Log in to a site. A real browser window opens at that URL with a capture bar along its bottom.
3. Do the login yourself — password, 2FA, SSO, CAPTCHA all work, because a human does them. Tick remember me or trust this device so the session lasts.
4. The bar’s button reads Capture now while you are on the sign-in form, and flips to Save session once a sign-in is detected. Press it when you are all the way in — after any code or phone prompt.

The window closes on save, or after a five-minute timeout if the login never completes. Nothing is saved until you press the button.

Magpie never types, clicks or solves anything in that window. Each capture starts from a fresh, logged-out profile.

What is stored

The session’s cookies — sealed, either in the OS keychain or as an encrypted sessions.json beside config.yaml, whichever this desktop offers. The Sessions table lists each one — domain, status, capture time, last use — and the cookie count at capture. Values are never displayed, never decrypted for the UI, and never leave the machine.

Delete a row and the session is gone. There is no cloud copy, because there is no cloud.

Using sessions

A scrape whose URL matches a captured domain attaches the session automatically — the address bar shows a session ✓ badge so you know before you run it. Research runs can read specific domains with your login: pick them under Read with my login in the Advanced drawer. The first use of each domain asks for a one-time acknowledgement of the terms — read-only, one page every 10 seconds per site, pages stored on this machine, each site’s terms apply to you.

When they expire

Sites end sessions; Magpie notices for you. The Sessions row shows an expired badge once an attached scrape hits an auth wall — the same scrape shows Session expired for <host> with a Re-login button that opens a fresh capture window. An expired login used by research reads like a logged-out visit; re-login first for full content.

For one-off needs, the Scrape screen’s Advanced drawer still takes Cookies by hand (a=b; c=d), and CDP endpoint can drive a browser you already logged into. Sessions are the durable version of both.