A source-linked AI reading workspace — every highlight, note, screenshot, and answer stays pinned to the exact spot it came from. Private, portable, yours.
Select text for a quick Highlight · Note · Ask AI popover, drop a point comment anywhere, or screenshot a figure/equation to ask about it directly.

Open an image-only PDF and PairedX notices there's no selectable text, then offers one-tap OCR that runs entirely in your browser — your file is never uploaded. It rebuilds a real text layer, so search, highlights, and the AI all work like a normal PDF.

A connector line joins each card in the panel to the highlighted passage or captured figure it came from — click a note and the reader jumps right to it.

Ask about any selection; the answer is saved as a note linked to that passage, with rendered math & code and a provenance panel showing exactly what it used.

For deeper questions the AI reads other pages, searches the document, scans the outline, or searches the web — and you can open “the agent's work” to see each call.

Screenshot any figure or equation and ask about it. The AI can also generate an image or a text/ASCII diagram to explain a concept, grounded in the paper.

Check any message — say the AI's summary instead of your question — to show it in full on the collapsed card. Plus tags, resolve, filters, search, and sort.

Home, Recents, Starred, and Trash. Star, trash, restore, or delete — the same PDF opened again is recognized by content (SHA-256) so its notes just re-attach.

OpenRouter (default) or any OpenAI-compatible endpoint. Your key stays in your browser and is sent per-request — never saved on a server. Toggle visuals and web search.

Edit the system prompts — even the agent's tool descriptions — under Settings → Templates, then export or import your whole set as JSON.

Everything stays in your browser. Optionally sync a portable .notes.json to a folder (Chrome/Edge) so notes travel with your PDFs — great for Drive and other devices.

The full list, grouped — every one of these ships today.