Compare

Pick two sides

1. Open Compare. Paste the URL both versions came from and press Load sides.
2. Each side — left and right — is a picker with two segments:
   • Runs — the scrape runs that checked this page in, newest first, labelled Run <id> · date, time.
   • Snapshots — every stored check-in of the URL, newest first, labelled Snapshot · date, time.
3. Pick one side each and the diff renders.

A scrape checks the page in when it finishes; a watch check files a snapshot on its schedule; research reads file their copies too. That is why the pickers mix kinds: the two sides of a diff are any two moments the page was stored, whatever fetched it. The lists reach the 500 newest runs and the 200 newest snapshots of the URL — a watched page’s older history stays reachable through the snapshot list, not by scrolling runs.

Both sides must come from one URL — a cross-URL diff is refused. Identical sides are not an error: the diff comes back empty.

The Watch screen’s drawer deep-links here: Open in Compare fills the URL and its newest snapshot.

Read the diff

The diff is word-level text: - prefixes the words a side lost, + the words it gained, one change per line. The pane headers name each side’s label, URL and check-in time, and both sides’ full Markdown stays readable above the diff for context.

Copy puts the diff text on the clipboard.

Export the diff

Export… opens the native save dialog and writes the same text as a .txt file, named after the left side’s timestamp. Magpie refuses a target inside its own data or config folders — the export can’t clobber the store, the records or the settings.

The CLI prints the same shape: magpie diff <url> --against snapshot.md (CLI reference). What lands in the file is covered on Export formats.