# About AI Canvas

> What AI Canvas is, how the canvas and its agent work, and how to fork the template.

AI Canvas is an open-source, forkable Next.js template for building your own AI design tool — a Canva, Figma, or tldraw alternative where the canvas can be driven in plain language. Type what you want and an agent places text, shapes, generated images, or live interactive HTML directly onto an infinite canvas, where you can keep editing it by hand.

The editor is a full canvas, not a chat window: zoom, pan, multi-select, layers, transform handles, keyboard shortcuts, and per-block controls for fonts, colors, borders, shadows, opacity, radius, and rotation. The agent sees what you see — every request ships a snapshot of the canvas (or just your selection) plus a structured summary of the blocks on it, so instructions like “make this headline bolder” or “add a pricing card next to the selected frame” land where you mean.

Under the hood it runs a single eve agent (Vercel's open-source agent framework) on AI SDK 7, calling models through the Vercel AI Gateway. Five tools cover the whole surface: generate a text block, a frame, or an AI image, and build or update a live HTML block. There is no database and no login — canvas state lives in your browser, and the only backend is the agent.

AI Canvas is built and maintained by Kaiyu Hsu. The hosted demo is a showcase; the point is the source, which you can fork, rebrand, and extend with your own block types and tools.

## Using the template

- [Source on GitHub](https://github.com/kyh/ai-canvas): fork it, or start a new repo from the template
- [AGENTS.md](https://github.com/kyh/ai-canvas/blob/main/AGENTS.md): the runnable setup guide for coding agents: install, one env var, `pnpm dev`
- [Vercel AI Gateway](https://vercel.com/docs/ai-gateway): where the `vck_…` key the demo asks for comes from
- **License**: MIT — use it in personal and commercial projects

---

[AI Canvas](https://canvas.kyh.io) · [All pages](https://canvas.kyh.io/sitemap.xml) · [llms.txt](https://canvas.kyh.io/llms.txt)