Pen

About

Pen is an agentic canvas for building bold ideas. You can design with agents directly on the canvas, or refine your work manually. Bring any model to the canvas, including Claude, OpenAI, Grok, Qwen, Gemini, Kimi, Deepseek, and Minimax, using a ChatGPT or Claude subscription, API keys, or MCP, WebMCP, or CLI. You can also connect agents such as Claude Code, Codex, Cursor, and others through MCP, or design headlessly with the Pen CLI.

Pen helps you start from existing work. You can import any website or app as editable layers, either by pasting from Chrome with the Chrome extension, using the built-in browser, or importing Figma .fig files. Its tools include a pen tool for drawing paths and pasting SVG code, shaders, gradients, variables, reusable components, component slots, UI libraries, masking, background removal, image cropping, and vectorizing images. Layers can be snapped, measured, grouped, and scaled, and the app supports familiar Figma and Sketch shortcuts.

When your design is ready, you can export any frame as PDF, PNG, JPEG, or WEBP, or export it as HTML and CSS/Tailwind. You can also share a link so anyone can open the latest version in their browser. Pen uses the .pen open design format, which is a JSON file that agents can work with natively. The app is built for performance, using a custom WebGL pipeline to keep thousands of layers smooth while you pan, zoom, and edit.