AI Co-Edit — Cursor for your notes.

Three layers of co-author inside a single editor — Reflect as you type, Comment on a selection, Co-Edit through a drawer chat. Each one knows what the others know, and all three read your knowledge base before suggesting a single word.

Co-author layers: 3
Block-level operations: 6
Selection actions: 7

AI Co-Edit

Experiment 03 · The Lab

Switch the layer. Watch the editor change shape.

A small note, three coordinated co-author surfaces. Pick a layer above the editor — the AI shows up exactly where that layer lives, and nowhere else.

Q3 launch — narrative draft 2 min ago · 4 backlinks · plan

Reflect · inline

What we are launching in Q3

A single editor for notes that knows what you know — and helps you write what you have not yet written.

The launch combines three months of focused engineering work with feedback from the closed beta. Customers will get a more mature product than what we shipped in Q2, with substantially fewer rough edges and a real pricing page.

Pricing follows the standard SaaS pattern: a free tier that lets you actually try the product, a Pro tier at fifteen dollars a month for serious users, and a Team tier for shared workspaces with admin controls. Annual plans are discounted.

Now showing: Reflect — the inline layer. Pause for a beat at any cursor position and the AI offers what comes next, in your voice. Tab accepts. Anything else dismisses.

Free includes Reflect on every note, with light daily limits. Pro unlocks Comment, Co-Edit and full knowledge-base context → See Pro

All three · how AI Co-Edit composes

Three layers, one editor.

Each layer answers a different question about when the AI should show up. Together they cover the full space — passive suggestion, active selection, full conversation — without overlap.

Layer I · Reflect inline

Inline ghost text

When you pause, the AI offers what you might say next — never the next paragraph, never the next idea, just the natural completion of the thought you have already started.

Pause-triggered Suggestions appear after a brief stop, never mid-thought. Type any character and the suggestion vanishes — the cursor is yours, the suggestion is optional.

Block-aware Different completions for headings, bullets, code, callouts. The model is told what kind of block the cursor sits in, and what kind of language belongs there.

Three-strike guard Three rejections in a row and Reflect goes quiet for the rest of the session. The signal is clear: not now.

Layer II · Comment selection

Bubble toolbar

Highlight any text and a bubble offers seven micro-actions — the writer's equivalent of margin notes that carry out the rewrite for you.

Fix Grammar, typo, clumsy structure — silent corrections, nothing else changes.

Rewrite Same idea, different wording. Useful when a sentence is right but reads wrong.

Tone Shift between academic, professional, casual, technical, creative or journalistic registers.

Summarize · Key points · To list · To table Four shape transforms — collapse a paragraph, extract a bullet list, restructure into a comparison table. Inline, no drawer needed.

Layer III · Co-Edit ⌘ ⇧ A

Drawer chat

Open the drawer and talk to the AI like a writing partner. It edits across blocks, surfaces relevant notes from your knowledge base, and waits for your nod before anything commits.

Update · replace · move Three operations that change a block in place — rewrite the contents, swap it for something new, or relocate it elsewhere in the note.

Insert before · insert after Two operations for adding fresh material — a new heading, a missing example, a transition paragraph that the draft was begging for.

Remove The sixth operation — delete a block when the AI agrees it should not be there. With pending state, removal is reviewed before it lands.

Knowledge-base context Before answering, Co-Edit pulls the relevant notes from your workspace through MCP — every reply is grounded in what you already know.

AI writing tools live somewhere else —
and forget your work the moment you close the tab.

Notion AI opens a panel beside the page. Jasper is a separate app. Quillbot replaces what you wrote with something else. Google Docs Gemini sees only the document in front of it.

They all have the same blind spot: they do not know what else you have written. The plan you drafted last week, the decision you made last quarter, the pattern across your last twenty meetings — invisible to the AI helping you write the next sentence.

The cost: You become the connective tissue. Copy a paragraph from your notes. Paste it into the AI window. Wait. Paste the result back. Edit. Re-format. The AI is helping someone — but it is not helping you.
The fix: Put the AI inside the editor. Let it read everything else you have written. Let it suggest, and let you approve.

The lineage

From collaborative editing to co-author.

AI Co-Edit did not appear from nowhere. Five moments — across human-computer interaction, office software, and AI editors — taught it how to share a document.

1968
Engelbart — The Mother of All Demos Live collaborative document editing, shown for the first time. The HCI assumption that two minds could share a single document was set that day, in a single 90-minute demo at Stanford. SRI · Stanford
1997
Microsoft Office — Clippy The first inline editor assistant to reach a billion screens — and to teach an entire industry what not to do. The instinct was right, the timing was wrong, the model wasn't ready. Microsoft Office 97 · cautionary
2009
Grammarly — selection-based AI Highlight a sentence, get a rewrite. The selection-as-target pattern that the entire bubble-toolbar generation inherited — first widely adopted in prose, before LLMs made it interesting. Browser extension · production
2023
Cursor — IDE-native AI co-author A forked VS Code that put the AI inside the editor instead of next to it — chat plus per-block edits, with the codebase as live context. The pattern AI Co-Edit ports to prose. Anysphere · March 2023
2026
Knovya AI Co-Edit Three coordinated layers — Reflect, Comment, Co-Edit — inside a BlockNote editor, every layer aware of the rest of your notes. The first co-author that lives where you write. Knovya · production

First of its kind

Nobody else co-edits inside a knowledge base.

Every AI writing tool either lives in a sidebar or in a separate app. The ones that do live in the editor — Cursor, GitHub Copilot — are built for code. AI Co-Edit is the first to bring that pattern to prose, with three coordinated layers and the rest of your notes as context.

ChatGPT Canvas chat-only · no KB
Notion AI sidebar · KB-blind
Google Docs Gemini document-bound
Jasper separate app
Quillbot paraphrase only
Cursor code IDE
★ Knovya 3 layers · KB-aware

Surfaces

Four small surfaces, one continuous editor.

Each layer takes up exactly the screen real estate it needs — no more. The drawer opens only when you ask; the bubble only when you select; the ghost only when you pause.

Co-Edit drawer ⌘⇧A

A side panel that opens with a keyboard shortcut. Talk to the AI, watch it pull context from your other notes, then accept or reject every block-level change it proposes.

Selection bubble comment

Highlight any text and a small bubble floats above it. Seven micro-actions, all instant — no drawer, no chat thread, just the rewrite back where the selection was.

Inline ghost reflect

Pause for a beat, the next words appear in lighter ink. Tab accepts, Esc dismisses, typing anything else replaces them. The cursor never leaves the line you are writing.

Per-block diff approve

Every Co-Edit operation lands as a pending diff — old block in red, new block in green, accept and reject sit one click away. Nothing commits without your nod.

Bonded with

AI Co-Edit composes with the rest of Group I.

MCP The protocol that gives the drawer access to your full workspace as live context.

AI Memory Shared brain across every AI you use — what Co-Edit reads when it answers.

Transforms Doc-level one-shots — the bubble's seven actions, scaled to the whole note.

Skills Composable AI workflows. Co-Edit becomes the surface where skills run.

Frequently asked

A few honest answers.

What is the best AI writing assistant for notes in 2026?

AI Co-Edit is a three-layer AI co-author built into the Knovya editor. Reflect suggests your next sentence as you type, Comment rewrites a selection in seven different ways, and Co-Edit opens a drawer chat that edits your note block by block. It reads from your knowledge base before suggesting anything, and every change is reviewable before it commits.

How is AI Co-Edit different from Notion AI or Cursor?

Notion AI lives in a panel separate from the document. Cursor is built for code, not prose. AI Co-Edit unifies both patterns in a single editor: three coordinated layers, native to BlockNote blocks, with the rest of your knowledge base as live context — not a separate window, not a sidebar wizard.

What are the three layers — Comment, Co-Edit, Reflect?

Reflect is the inline layer — a ghost suggestion that appears after a short pause, dismissed by typing or accepted with Tab. Comment is the selection layer — highlight any text and a bubble offers seven micro-actions like rewrite, summarize, or change tone. Co-Edit is the drawer layer — a side panel where you talk to the AI and it edits the note across multiple blocks at once.

Does AI Co-Edit read my other notes?

On Pro, yes. The Co-Edit drawer can search your full workspace through Knovya's MCP layer — every relevant note becomes context before the AI writes a single word. Reflect and Comment also incorporate the surrounding note. End-to-end encrypted notes are excluded automatically.

Is AI Co-Edit free?

Reflect — the inline ghost layer — is included on every plan with light daily limits. Comment and Co-Edit unlock on Pro, with knowledge-base context and the full set of edit operations. Both tiers count against the same monthly AI credit pool.

Can I undo what the AI wrote?

Nothing the AI suggests is committed until you approve it. Every edit is shown as a pending change you can accept or reject — block by block. The original note is held in a snapshot, so a single click reverts everything. The principle is simple: AI suggests, you approve.

What AI models power AI Co-Edit?

AI Co-Edit routes between fast and quality tiers depending on the action. Inline suggestions and short transforms use a fast model for low-latency response; drawer chat and long-form rewrites use a quality model. Knovya manages the routing — you write, the right model answers.

Try the three layers in your own notes.

Reflect on every plan from day one. Comment and Co-Edit unlock the full co-author surface on Pro.

Start free See Pro

★ element 03 · Group I — AI