v1.3 — document chat for every file, single files, and web apps

MindMax v1.3 removes the friction between your documents and AI. Until now, asking questions meant having an entire workspace folder open and explicitly pulling a document into the right-hand chat. v1.3 makes every document immediately conversational, lets you open standalone files without a project, adds instant actions to every document image, and introduces three lightweight web apps so you can write, think and read in any browser.

What is in it

• Document chat for every file. The right panel now belongs directly to whichever document you have open — Markdown notes, PDFs, EPUBs, slides, data files, canvases, and images. Ask "what does this section argue?" or "tighten this paragraph", and the assistant answers with full awareness of that specific file. No need to @ files or manually add chips.
• Launchpad as the Workspace Agent. When you want to ask across your entire library rather than one document, the Launchpad and the left-rail agent provide a dedicated conversation space. The division is clean: document questions stay beside the document; broad inquiries across all notes happen on the Launchpad.
• Open single files without a workspace. You no longer need to create or pick a workspace folder just to read a paper or review a draft. Open any Markdown, PDF, EPUB or PPTX file directly from Finder or Explorer, or drag it onto the welcome screen. Reading guides, OCR, citations and document chat work immediately.
• Image hover toolbar. Move your cursor over any picture in the Markdown editor or in reading view (including figures in PDFs and slides) to reveal an action toolbar. Ask the AI about the image with one click, zoom in, convert it to a note, or transcribe it into text inserted below the picture as a tracked revision. You can also send the image straight to a canvas.
• MindMax Write online editor. A zero-install browser document editor at mindmax.me/write. Paste AI responses from ChatGPT or Claude without breaking math formulas, tables, or Mermaid diagrams. Edit with inline AI and tracked revisions, then export cleanly to native Word (.docx with real OMML equations), PDF, or image cards.
• MindMax Think online canvas. A structured AI thinking canvas at mindmax.me/think that applies proven consulting and analytical methods (MECE, premise audits, SCAMPER, steelman arguments). Ideas expand as connected cards that converge into outlines, which you can hand off to Write with a single click.
• MindMax Read in the browser. A web reader at mindmax.me/read that unpacks chapter structures, provides reading guides, and answers questions with verifiable page citations.
• Shareable card pages. Publish long notes or briefs into responsive, swipeable card pages hosted at cards.mindmax.me. Readers get a clean mobile reading deck where every card links back to its verified source quote, while links remain permanent even after you update edits.
• Two-column CJK paper reading. Academic papers in two-column layouts now read column-by-column without mixing lines across columns. Footnotes, figures, and inline formulas are parsed faithfully without being mistaken for tables, accompanied by an AI math recognition shortcut.

Known rough edges

• Single-file reviews and backlinks. Opening a file outside a workspace disables workspace-wide backlinks and mentions, since there is no surrounding index of files. If you want full cross-linking, open its folder.
• Web app offline storage. The web versions store drafts locally in your browser's IndexedDB. Clearing browser data or using private browsing windows removes local drafts unless you export them or log in.
• Image toolbars on touch devices. On tablets and touchscreens, image hover toolbars appear on tap rather than hover, which can take an extra tap to dismiss.

Where to start

Try the new document chat by opening any note or PDF and typing in the right panel. To use MindMax on the web without installing, try MindMax Write or MindMax Think. Already using the desktop app: the update will download automatically — restart when prompted to switch to v1.3.0.