Release notes and roadmap
Roadmap
Pro on the web
Monthly and annual Pro, eligible 14-day trials, and founder pricing.
Answers that cite their sources
Answers and Briefs link to the saved items they used.
Chrome extension
Save pages, selections, and whole windows of tabs.
Organizations
Invite people and manage members. Each person’s library stays personal.
Sign in with Google and Apple
Sign in, or link an existing account, in one step.
One-click Notion connection
Connect Notion without pasting a token.
Second on the App Store
iPhone, iPad, and Mac apps, with the Safari extension. The public beta is open on TestFlight.
Obsidian community plugin
Install two-way vault sync from Obsidian’s plugin browser.
Action history
One list of every change made by you, ChatGPT, or Claude, with undo for each.
Public REST API
Capture, retrieve, organize, and export over REST with personal API tokens.
Merge duplicates
Ask ChatGPT or Claude to merge duplicate items into one, keeping every source.
Direct handoff between AI apps
Start in ChatGPT, then open Claude and say “continue”: it picks up the handoff on its own.
Per-chat folder access
Choose which projects each individual chat can read, not just each connected app.
Releases
Web 1.12.6Fixes
Fixes and improvements
- Successful AI Search indexing now clears any stale
index_failedmarker on the revision (healed rows no longer display an error that no longer applies). - Search indexing now records the real error message (and logs it to Workers Logs) instead of a flat
index_failed. Uploads are accept-and-go — Cloudflare finalizes indexing asynchronously — and a failed upload is retried through the existing retry-ingestion path on demand.
Web 1.12.5Fixes
Fixes and improvements
- Fix AI Search indexing failing on every capture: uploads are now accepted without polling for finalization (
uploadAndPoll's fixed timeout turned the platform's slow finalization into fakeindex_failedon 400+ items — the documents indexed fine in the background). Also fixes re-running ingestion for an already-live revision crashing on the active-ready constraint.
Web 1.12.4Fixes
Fixes and improvements
- Re-saving an already-captured URL with page-extracted content now heals it: when the existing item's latest revision failed or never got content, the provided Markdown becomes a new revision and re-indexes (previously dedupe dropped the content entirely). Captures that already have ready content are untouched.
Web 1.12.3Fixes
Fixes and improvements
- Canonical Markdown normalization no longer skips consecutive raw HTML nodes (splice-during-visit now revisits the shifted sibling), so captured content can't leak stray markup.
- URL captures accept page-extracted Markdown (
contentMarkdown) so browser-extension saves carry real content from the user's session; the ingestion pipeline skips rendering when a revision already has content.
Web 1.12.2Fixes
Fixes and improvements
- Harden PDF capture ingestion: converter output that arrives as raw HTML (table-dense documents like service manuals) is salvaged to plain text instead of failing as empty, PDF responses are validated by magic bytes so bot-wall stubs can't poison conversions, and the size cap rises to 25 MB for large manuals and brochures.