NewShared links look like your board

Your app looks AI‑generated.

Krit is where your agent designs: the AI look stripped, your brand captured as rules, every screen shipped to your repo looking like screen one.

Pro is $1 for 3 days, then $9.99/mo

Untitled
Any agent
Signed-in user
−55%+
Claude Code · acme/app · main
→ Add a follow-up
claude-opus-5

Designed on Krit

Real work, drawn on the canvas.

Every frame was prompted and shipped on Krit: live artboards and real canvas sessions, never screenshots. Hover to pause, click to open one.

Every frame was prompted and shipped on Krit: live artboards and real canvas sessions, never screenshots. Tap any one to open it full size.

Connect, Linear-grade
Card game lobby
Server console, redrawn
Retrieval workbench
Proof tracker, glass
Train tray, live
Court booking
Swiss data report
Proof dial
Liquid slider
ATELIER portfolio
Sleep recovery
Weekly digest
Proof field
Pixel sticker

Your brand, every screen

One product. Every screen.

Harbor, nineteen screens later. Krit captured the visual language once, and your agent has drawn against it ever since.

Visual language
ink1B1A18paperF9F7F2accentE0704Fradius12typeInter · 500densitycompact
screen 01 · home
Harbor
Good morning, Sarah.
What do you want to know?
Ask anything about your workspace
+SourcesThink↑
Q3 tool spendWho owns Datadog?Draft the standup
screen 04 · thread
‹Q3 tool spend
Today
Is billing connected yet?

Yes. Stripe synced 20 minutes ago.

What did we spend on tools in Q3?

You spent $4,820 on tools in Q3, up 9% from Q2.

Figma$1,860
Vercel$1,240
Linear$1,020
Datadog$700
Sourcesbilling.csvq3-report.pdf
Which one grew fastest?

Datadog, up 34% on new host volume. Everything else held flat.

Anything we can cut?

Two Linear seats have been idle for 60 days. Dropping them saves $340 a quarter.

Sourcesseats.csv
Ask a follow-up↑
screen 09 · settings
‹Settings

Workspace

Weekly digest
Monday, 9:00 AM
Cite sources
Show files in every answer
Share threads
Off for the whole team
Default model
Harbor 2 · balanced
Change

Notifications

Mentions
Push and email
Thread replies
Push only

Plan

Team · 12 seats
Renews 1 Oct
Manage
Billing email
sarah@harbor.app
Edit
Invoices
Last paid 1 Sep
View
screen 19 · connect
‹Connect a source

Connected

S
Stripe
Revenue and invoices
Connected
N
Notion
Docs and specs
Connected

Available

G
GitHub
Issues and releases
Connect
L
Linear
Roadmap and cycles
Connect
F
Figma
Files and comments
Connect
S
Slack
Threads and channels
Connect
D
Drive
Docs and spreadsheets
Connect
Z
Zendesk
Tickets and replies
Connect
Continue
Same header. Same palette. Same components. Screen 19 was drawn from the same rules as screen 01, and the rules live in your repo.

One canvas, end to end

It doesn't stop at a mockup.

The same agent that drew your design can test it, animate it, and hand you the code, without leaving the canvas.

Usability testing

Find the confusing screen before your users do.

Personas read the screen, try to finish the task, and narrate where they hesitate. Each finding becomes a prompt your agent fixes on the canvas, before anything merges.

Harbor · Q3 thread
‹Q3 tool spend
What did we spend on tools in Q3?

You spent $4,820 on tools in Q3, up 9% from Q2.

Figma$1,860
Vercel$1,240
Linear$1,020
Datadog$700
Sourcesbilling.csvq3-report.pdf
Ask a follow-up↑
HRight, Q3 spend.
Usability run0/3

Task: check what we spent on tools in Q3

HHurried
FFirst-time
PPower user
Findings0
fix: make source pills read as tappable
copied as prompt · your agent ships it in PR #47
Harbor · Q3 answer

You spent $4,820 on tools in Q3, up 9% from Q2.

Figma$1,860
Vercel$1,240
Linear$1,020
Datadog$700
Sourcesbilling.csvq3-report.pdf
SpendSummary212 × 36
ReactNext.jsVueHTMLTailwind
SpendSummary.tsx10 lines
1type Props = { total: number; delta: number };23export function SpendSummary(p: Props) {4  return (5    <p className="spend-summary">6      You spent <b>{fmt(p.total)}</b> on tools7      in Q3, up {p.delta}% from Q2.8    </p>9  );10}

Export to code

Hand it off as code, in whatever your stack speaks.

Select anything on the canvas and export clean, framework-native source. No data-cv-id, no canvas runtime, no redraw.

Launch assets

The launch video ships with the feature.

Ask in plain language. The agent composes the entrance, the camera, the easing, and Krit renders a real MP4. The demo for your launch post, the changelog GIF, the hero video, cut from the product you just built.

Harbor · launch reel

You spent $4,820 on tools in Q3, up 9% from Q2.

Figma$1,860
Vercel$1,240
Linear$1,020
Datadog$700
Claude Code · harbor/app
0.0s / 4.0slaunch.mp4 · 1080p
camerarows
0s1s2s3s4s
slide-up · easeOut
stagger · 110ms

Your agent, multiplied

Your agent brings its other tools with it.

Krit sits alongside the tools your agent already holds, and anything it can read becomes a design move.

A product that fixes itself.

Your agent finds the step people abandon, redraws it on the canvas, and opens the pull request.

Checked before merge

Every branch measured against the visual language you captured.

The states you skipped

Empty, error, loading, and every edge the happy path forgot.

One change, everywhere

Change the rules once and every screen follows.

Live, not a picture

Click through a frame the moment it exists. No prototype to wire up.

Every width, both themes

Phone through desktop, light and dark, from the one set of rules.

Sound and haptics

Ask for a tap that clicks and a confirm you can feel.

No integration to wire up, no setting to turn on.

Pricing · Pro

$9.99
per
month
$1 for 3 days. One plan, everything included.

For anyone building with an AI agent. No seats, no usage limits.

Start 3 days for $1

Then $9.99/month. Cancel during the trial and you never pay the monthly price.

What's included
Unlimited designs & canvases
Works with any AI agent over MCP
AI usability testing with personas
Storyboards, motion & video export
Exports clean HTML & Tailwind, ready for your stack

FAQ

Questions, answered.

Which AI agents work with Krit?
Any agent that speaks MCP. Claude Code, Cursor, Windsurf, Cline, Codex, and more. Connect once and the agent draws on the canvas the way a designer would.
Is the exported code actually production-ready?
Yes. Krit writes real HTML and Tailwind from the start, so the export is the exact markup you designed. No data-cv-id, no canvas runtime, no redraw. React, Next.js, and Vue exports wrap that same markup as framework-native components. HTML and Tailwind are the most battle-tested path.
Do I need design experience?
No. Krit was built for founders and product engineers, not designers. Describe what you want; the agent handles layout, type, color, and motion. You steer with words.
Is there a trial?
Yes, 3 days for $1. Cancel before it ends and that $1 is all you pay; otherwise it becomes the $9.99/month Pro plan and the first monthly payment is taken the day the trial ends.
Can I cancel anytime?
Yes. The $9.99 Pro plan is monthly with no contract. Cancel from your account and keep every design and export you've made.

From the blog

Notes from building this way.