Recall Release Notes: May 15, 2026 - Post-launch improvements, fixes, and bug bashes.
It's been a month since the launch of Recall 2.0. We've taken some time to rest and recover, while bashing bugs, smoothing out the experience, and figuring out what's next. Help shape what's next for Recall - check out our "What's Coming" feature to share your input.
New features
Bulk PDF uploads (Plus and Max). Upload multiple PDFs at once. On Max, AI actions run automatically. On Plus, bulk tag manually and start chatting.
Share PDFs from your phone. Use the share sheet to send PDFs directly to Recall.
iOS swipe-back gesture and general card navigation polish.
Right-click to open cards in the tag tree, chats, and citations.
Updated MCP and API docs. See docs.
Improvements
Smarter AI actions on notebooks which now use the full note, including attached files.
More robust TikTok scraping in the browser extension.
Cleaner file uploads with scroll indicators and a tidier footer.
Smoother mobile navigation with hidden scrollbars.
More reliable file previews in chat.
Fixes
X posts now save on mobile. Full list of supported content.
Connections for PDFs without a source URL.
Quiz infinite loading resolved.
Tag initialization more reliable across app and extension.
Back-swipe from cards no longer hits the login screen.
iPad pointer and touch handling fixed.
Clickable links restored: Command/control-click opens cards in a new tab.
Mobile styling fixes, including card corners, editor safe area, and hero image padding.
Old iPhone crash resolved. No more crashes on iOS Safari below 16.4. Unsupported browsers fall back to CommonMark.
What's next
A smoother, more intuitive UI. We're going through the app end-to-end to make navigation, interactions, and visual hierarchy feel more natural. Less friction, fewer dead ends, more "where I expected it to be."
A better help flow. Easier to find answers when you're stuck, with more discoverable docs, in-app guidance, and faster paths to support.
Improved search and better export. Two of the most-requested basics. Search is getting smarter and more forgiving. Export is getting more flexible so your knowledge base stays portable.
Browser Extension Update: We've decided to move forward with the original version of the browser extension, which was a pop-up vs side panel. A few things to note:
It provides support across more browsers.
Enables you to create recall cards on multiple pages at the same time.
It will be locked by default and you can resize it.
Write API and a more scalable way to add bulk content to Recall: this will require some significant investment, so we will be making progress toward it and will come a little further down the line. It's on our radar and something we are working toward.
Coming soon: Public Recall profiles
We've seen an increase in shared cards, and it feels like the right moment to make the great content being created in Recall easier for others to find and learn from. The idea: share content publicly and get a Recall profile with a bio and a feed of your shared cards and collections. Curate collections like "Best resources on agentic AI" as a playlist of summaries. Others can discover, like, comment, save, and follow you for new posts. Think of it as the lightweight, Recall-native version of something like Medium: a place to share what you're already learning, without the effort of writing long-form or the friction of sign-ups for readers.
Check out the feature request and add your thoughts.
Replies
Strong release notes. The bit that jumps out to me is “AI actions use the full note, including attached files” paired with better search/export on the roadmap.
For personal knowledge products, trust usually comes from two things: can I trace what the AI used, and can I move my knowledge somewhere else if I need to? One suggestion: when an AI action pulls from PDFs/cards/notes, show a compact “used these sources” trail with freshness or confidence cues. That makes the knowledge base feel less like a black box and more like an inspectable thinking partner.
Recall
@jim_jeffers You raise a great point. We definitely always link our sources, especially based on the cards they come from. We can do a better job of explicitly stating whether it was the notebook, the reader, or the attachment.
@sankari_nair That source-type label would help a lot. The subtle trust issue is usually not "is there a citation?" but "what kind of thing is the AI leaning on right now?"
For example, notebook synthesis, reader selection, attachment text, and user-written note all deserve slightly different confidence. Even a compact receipt like Source: 2 cards + 1 attachment + notebook note would make the answer easier to review without turning the UI into an audit log.