It reads, not writes
No blank-page takeover. No prompt theater. No trying to co-author your manuscript. KRW starts after a draft exists and helps you see it more clearly.
Reading-first revision environment
An AI that learns how to read your book.
Most AI writing tools want to join the writing. KRW starts after the draft exists. It reads your manuscript like an attentive editor, surfaces the few passages worth a second look, and remembers what you tell it. The more you use it, the quieter it gets.
- Fewer notes that actually matter
- Every decision teaches the system your intent
- Your manuscript stays in your browser
Chapter 7
3 findings worth attention
Mara kept the lantern low as she crossed the ruined nave. The glass underfoot gave back almost no sound. At the altar she stopped, not because of the body, but because of the impossible neatness of it. Someone had arranged the dead boy's hands.
The first time she saw Daniel after the fire, he smiled in It made her distrust the kindness before she trusted the grief.
Outside, bells were already starting.
Why this exists
Every AI writing tool wants to help you write. This one just reads.
There's no shortage of drafting copilots, grammar checkers, and story bible platforms. KRW doesn't do any of that. It reads your manuscript the way an attentive editor would, notices what's worth a second look, and learns what you meant to do on purpose.
Your intent wins
Tell the system why something's deliberate and it remembers. That one decision shapes how every future scan reads your book.
It should feel like reading a book
Not like operating software. One panel open at a time. No sidebars. No layout shifts. The manuscript is the interface.
Interactive demo
Notice, decide, teach, continue.
KRW isn't scoring your text. It's learning how your manuscript should be read. Every time you dismiss, explain, or defer, the next scan gets smarter. That's the whole loop.
Chapter 7 scan
Mara kept the lantern low as she crossed the ruined nave. The glass underfoot gave back almost no sound.
The first time she saw Daniel after the fire, he smiled in exactly the same way and she distrusted the kindness before she trusted the grief.
Outside, bells were already starting. Not an alarm. A welcome.
Teach once
The highlighted line is not a mistake. It is how this narrator judges people before the story proves her wrong.
Why this is intentional
This narrator makes fast moral judgments that get revised later.
intentional narrator sharpness
save to reading memory
Next pass
3
findings before
1
finding after
Voice noise reduced around intentional sharpness
Pacing note still surfaced for this chapter ending
Author guidance reused automatically
Before: AI flags a passage
Click a response to see what happens
She kept walking past the chapel entrance, where someone had left candles burning in glass jars.
She smiled the way people do when they have already forgiven themselves. The warmth in her expression was aimed at no one in particular.
It was a small chapel. The kind of place that always smelled like stone.
Voice: This sharpens character judgment while potentially flattening ambiguity. Worth attention, but not necessarily wrong.
Dictionary: "This narrator makes fast moral judgments, intentional"
What happens next
Choose a response
Every response teaches the system something different.
Click Explain why, Got it, or Later on the left. Each one shapes how the AI reads future passages. That's the kaizen loop: small acts of attention, compounding.
- Teach: the AI remembers your intent across every future scan
- Got it: the mark resolves and similar noise drops to almost nothing
- Later: the passage is saved for when you're ready to deal with it
What you actually get
Built for revision, not for show.
Everything here exists to keep you in the manuscript. Read, respond, move on.
Focused chapter scans
The AI reads each chapter and marks a handful of passages. Not hundreds. Just the ones worth your attention.
It actually learns
Dismiss a mark and the system reduces similar noise. Teach it why something's intentional and it remembers across every future scan.
dismiss→
learn→
adapt
Keyboard-native
Tab through marks, Enter to open, D to dismiss, T to teach, F to flag. You don't have to leave the text.
D
T
F
One panel at a time
No sidebars. No drawers. Notes expand inline, right where the mark is. Open one, the last one closes.
Full-manuscript pass
After the chapter-level scans, a cross-chapter pass catches the bigger patterns: continuity gaps, repeated motifs, pacing across the arc.
across chapters
Density controls
Too much voice feedback? Turn it off. Only care about pacing today? Show just that. You shape the pass to fit the revision you're doing.
on
off
on
Works on mobile
Tap for the table of contents, swipe between chapters. The reading experience carries over to smaller screens.
Export and backup
Save your marks, export a revised manuscript, or pull a DOCX with comments. Your editorial conversation doesn't disappear.
Who it's for
Fiction writers revising real drafts.
KRW fits the stretch between finishing a draft and sending it to an editor or beta reader. It's for the author who wants developmental-level feedback before the manuscript leaves their hands, and who cares about preserving voice as much as improving clarity.
- Draft 2 through 4 novels
- Indie authors doing their own revision passes
- Writers who want fewer notes, not more
- Authors who don't want AI flattening their prose
Not another AI writing app
- Not a blank-page drafting copilot
- Not a grammar checker chasing every sentence
- Not a sprawling story bible platform
- Not a co-author trying to take over the page
KRW's lane is narrower on purpose: a better first reader for revision.
Trust
Your manuscript doesn't leave your browser. Your keys don't touch our servers.
Writers don't just need capable AI. They need a product that feels safe, predictable, and easy to understand.
Local manuscript state
Your manuscript, your marks, your dictionary all live in your browser. Nothing gets stored on a server.
Bring your own key
You pick the AI provider. You add your own API key. It goes directly to the model, not through us.
Clear about what it's doing
You can always see what the AI noticed, what you taught it, and what it dismissed. There's no hidden logic.
Simple by design
No framework overhead. No app-shell bloat. It's fast because there's nothing in the way.
Under the hood
No build step. No framework. Just the parts that matter.
The architecture is intentionally minimal. Static frontend, a lightweight AI proxy, browser storage, and streaming responses. Nothing between you and your draft that doesn't need to be there.
Frontend
Static HTML, CSS, vanilla ES modules
AI layer
Streaming editorial reader with provider routing
Models
Gemini 2.5 Flash Lite and Claude Sonnet
State
Browser storage only, no server-side persistence
It's lean on purpose. No account system, no server-side manuscript storage, no framework overhead. The whole thing stays fast because there's almost nothing to slow it down.
Principles
What KRW cares about
FAQ
Common questions
Is this trying to replace editors or beta readers?
No. KRW gives you feedback before your pages go to other humans. It's a first reader, not the final authority. Think of it as the pass you do before the pass that matters.
Does KRW write for me?
No. It reads your manuscript and marks what's worth attention. You decide what to do about it. The goal is to help you see your own draft more clearly, not to rewrite it.
How is this different from other AI writing tools?
Most AI writing tools try to do everything. KRW stays narrow on purpose: it reads, it learns what you tell it, and it gets quieter over time. It's built for revision, not drafting.
Will it keep flagging choices I already explained?
That's the whole point of the teach loop. Explain an intentional choice once and the system carries that context into every future scan. It shouldn't flag the same thing twice.
Early access
Your draft doesn't need another AI trying to write it. It needs a better reader.
KRW is for authors who care about judgment, not just capability. Drop your email and we'll send you an access link.
No account. No password. We send you a link. Bookmark it, that's your access.