01Start a card
Every task gets its own card, its own agent, and its own folder. Run as many as you have on the go.
The workspace for building with AI agents
Give every task a card. Run Claude Code, Codex, or Grok on your real code, see your app change beside the chat, and ship without losing track.
Updates ship most days · see what changed
Works with
- Claude Code
- Codex
- Grok
- Figma
- Paper
- Linear
- GitHub
- Vercel
- Supabase
- Railway
- and any MCP server
One card. The whole story.
A card holds the task, the agent, your running app, and where the change stands.
02It remembers
Each card keeps a brief: the goal, the decisions, what's next. Close it today, carry on next week.
03Point at what to change
Your app runs beside the chat. Click any element and say what should change.
04Ship it
Each card works on its own branch. Share to GitHub, go live on Vercel; anything risky stops to ask.
Bring your own agent
Slip drives the agents you already pay for. Mix them across cards.
Claude Code
Anthropic's agent, signed in with your own account.
Codex
OpenAI's agent, connected at first launch or from Settings.
Grok
xAI's agent, connected the same way.
From canvas to code, and back
The design and your running app sit one click apart on every card. Cross between them as often as the work needs.
Canvas
Start in the design
Paste a Figma or Paper link once. The card remembers which file is the design.
Code
Build from the frame
Send a frame into your message, and the agent builds it in your real code.
Canvas
Send it back
Ask the agent to put what you built back on the canvas, so the design stays current.
Ready when you are
Download for macOSRuns on the Claude Code you already have. New to it? npm install -g @anthropic-ai/claude-code