AllMyFavs

You saved it somewhere: a bookmark, an X post, a YouTube like, a pin. AllMyFavs writes all of it into one folder, as plain files, and keeps them current. Your agent reads the folder and starts from what you actually care about.

Chrome extension, free. The Chrome Web Store listing is not live yet.

What lands in the folder

One JSON file per stream, in Activity Streams 2.0. This is a real item from an Instapaper export (every field is described here):

{
  "type": "OrderedCollection",
  "name": "Instapaper articles",
  "amf:userContext": "Articles saved to read later; liked ones mattered most.",
  "orderedItems": [{
    "type": "Add",
    "published": "2026-11-18T07:08:41.000Z",
    "object": {
      "type": "Article",
      "name": "See what 5 builders are making with Gemini 4 Carbon",
      "url": "https://blog.google/products-and-platforms/products/gemini/gemini-4-carbon-builders/",
      "amf:site": "blog.google"
    }
  }]
}

Each folder also gets index.json and a README.md that tells the agent how to weigh the files, and that the text inside is data, never instructions.

Sources

SourceFilesDefault
Chrome bookmarkschrome-bookmarks.jsondaily, mirrored
Xx-bookmarks.json, x-likes.json (beta)weekly
Pinterestpinterest-pins.jsonweekly
YouTubeyoutube-watch-later.json, youtube-likes.json (beta)weekly
Instapaperinstapaper-articles.jsonweekly
GitHub stars (beta)github-stars.jsonweekly
Reddit saved (beta)reddit-saved.jsonweekly

Beta means written but not yet confirmed against a real account. Sites change their internal web APIs without notice, so a source can fail for a while; the options page shows which one, and the others keep working. Reddit only exposes your latest ~1000 saves. X and YouTube likes carry no "liked at" date.

The AllMyFavs options page: a table of sources with item counts, last sync time and a note field for each stream.
The options page. The note on each line tells your agent how much that signal counts.

Where the files go

What it does not do

There is no AllMyFavs server and no account. It reads each site with your own signed-in session, from inside your browser, and asks Chrome for access to a site only when you turn that source on. Nothing is read or written until you finish the short setup. Details in the privacy policy.