Pin a comment. Get a redesign.

Click on your app and say what looks wrong. Your AI draws a better screen. Say yes, and it builds it.

  • Better screens, without hiring a designer

  • Always your brand, your colours, fonts and logo

  • No extra AI bill, it uses the AI you already pay for

Your agent writes the code in minutes. Good design is the part it still gets wrong.

It sees the whole flow

The board holds every screen of every journey. The agent designs the step between two screens, not one screen in a vacuum.

It designs in your brand

Every design renders with the tokens, fonts and logos from your corporate design page. The agent has no room to invent a colour.

It checks its own work

After building, it screenshots each screen of your running app and compares it with the design you approved, then fixes what differs.

The same journey, before and after one design round

Drag the handle. Left, the screenshots of what ships today. Right, the designs the agent proposed on the same cards.

Design mode: the same five onboarding cards showing the agent's designs, marked Approved.
Implementation mode: five onboarding cards of FitFuud showing screenshots of the shipped app.
Implementation what ships today Design what your agent proposed

FitFuud, the first product designed on Grafaris. One switch in the header flips every card between the two.

One round, up close

Four comments in, one sign-up out

Each pin names one mistake. The agent answers all four in a single new version, and the old one stays on the card to compare.

Screen sizes and languages

Every size, every language, one card

Add the sizes your product runs on, from a watch to a wide desktop, and the languages it speaks. Each card keeps its captures and designs per size, and the board switches with one click.

  • Phone, tablet, desktop, watch, or any size you type in
  • Two sizes side by side on the board
  • Your AI designs in the base language, then each translation

How it works

From comment to shipped screen

Click a step to see it. The steps move on when you stop scrolling.

1You

Pin the comment where it hurts

Click the exact spot on a screen and say what is wrong. Paste an image, mention a teammate, keep the thread.

2Your agent, put_design

Your agent answers in HTML

It reads the open comments and your brief over MCP, writes a new design version and replies on the thread.

3You

Approve one version

Versions never overwrite each other. Compare them on the card, ask for another round, approve the right one.

4Your agent, grafaris-implement

It builds it and checks itself

The implement skill walks the journey in a browser, screenshots every screen and compares it with the approved design.

Paused while you read
1. Three comments on one sign-up screen
2. Your AI sends the HTML, the board renders it
3. You approve B, A stays on the card
4. The skill checks the board again and again until your app matches

Corporate design

Your whole corporate design on every screen

Set up your corporate design once: colours, fonts, logos, icons, spacing and corners. Every design your AI makes follows all of it. On Premium, Grafaris itself wears it too.

Switch design

Logo Studio and Brand Studio

Your logo and your brand, next to your screens

Two studios sit on the same board as your journeys. Your AI works in them the same way: you comment, it answers.

Logo Studio

Logo Studio

Your AI draws logo concepts and new versions of the one you like.

  • Every variant, on dark and on light
  • See it in use on mockups
  • Download the finished files
Brand Studio

Brand Studio

Say what your brand should stand for. Your AI researches how people actually see it.

  • Brand identity map, yours plus the AI's additions
  • Brand image map with sources for every point
  • A match score between the two

Bring your own AI

Runs on the AI you already pay for

Grafaris has no AI of its own. The board and your comments live online. Your code and your AI stay on your computer.

  1. 1You comment on a screen on the board. Your AI picks it up.
  2. 2Your AI draws a new design and puts it on the board.
  3. 3You say yes. Your AI builds it into your code, and your app gets the new design.

14 days free on every plan

Pick your plan and start today

Every plan lets your AI design and build your screens. Pick by how many products and people you have.

Starter

Everything you need for one product.

  • Your AI redesigns any screen you comment on
  • Logo, colour and brand studios
  • 1 project with 5 journeys
  • Last 10 versions of each design
Start 14-day trial
Recommended

Premium

For several products, clients or brands.

  • Everything in Starter
  • Unlimited projects and journeys
  • Every version kept
  • 3 visitors who view and comment
  • Grafaris in your own brand
Start 14-day trial

Team

One board for your whole product team.

  • Everything in Premium
  • Members and roles
  • Threads, mentions and tasks
  • Free visitors, as many as you need
  • SSO at launch
Start 14-day trial

A card is needed to start. Cancel before day 14 and you pay nothing. Compare every plan

Before you start the trial

Questions people ask first

One short answer each. The detail sits on the pricing and privacy pages.

  • Tested with Claude Code and Codex
  • Web apps today
  • Servers in Germany, no AI of our own

Setup

Which AI agents work, and do I need a paid plan for them?

If you are comfortable in a terminal, use Claude Code or Codex. That is the connection we recommend, and both do the whole loop, including building into your code. Otherwise connect the Claude app, ChatGPT, DeutschlandGPT or Grok. Your AI plan must allow custom connectors: Claude Pro, Max, Team or Enterprise, or ChatGPT Plus, Pro, Business, Enterprise or Edu. Grafaris adds no AI cost on top.

Which apps work?

Web apps, for now. Native mobile and desktop apps are not supported yet.

How do my screens get onto the board?

Your agent opens your running app in a browser, walks each journey and uploads one screenshot per screen. You can also start from the URL of a live site.

Can a designer without a coding agent use it?

Yes, to comment, compare designs and approve. For new designs, someone on the project connects an AI, and the Claude app needs no terminal. Building a design into your code still takes a coding agent.

Can I import from or export to Figma?

No. Designs are HTML in your brand tokens, so your agent builds them straight into your code.

Data and cost

Do screenshots of my unreleased app leave my computer?

Only the screenshots you put on the board, and they are stored with your project on our servers in Germany. Your code and your agent stay on your computer.

Every processor and storage place: Datenschutz

Does my AI plan's usage go up?

Yes, a little. Each design round is an ordinary agent session on your own plan, like asking your agent to restyle a screen.

Plans

What happens when the trial ends, and can I cancel monthly?

Cancel before day 14 and you pay nothing. After that your plan bills each month and you can cancel for the end of any month. When it ends, your organisation turns read-only and nothing is deleted.

It's early access. What works today?

The whole loop works today. Prices, SSO and the audit log come at launch.

Journey board and commentsMCP for your agentDesign and implement skillsLogo and Brand StudioPrices at launchSSO and audit log at launch

Plan limits and billing: Pricing

Make your screens look great

Try Grafaris on your own product for 14 days. Your AI does the design work, you pick what ships.

Start 14-day trialCancel before day 14 and you pay nothing.