# A design canvas for GitHub Copilot

> Brilliant gives GitHub Copilot a real vector design canvas. Run one command and ask for a screen in VS Code: Copilot's agent creates and edits frames, auto layout, components and text as editable elements, reads what you changed by hand, and exports PNG, SVG, PDF, HTML or React.

Generated from `src/data/landingPages.js`, the same source the React page and its FAQ structured data read. Edit there, redeploy, and this mirror updates.

## How do I connect GitHub Copilot to Brilliant?

One command. It needs Node 20 or later and no API key.

```text
npx -y @brilliant-hq/mcp install vscode
```

1. **Run the command.** Paste it into your terminal. It adds Brilliant to VS Code, where Copilot's agent runs, as an MCP server.
2. **Ask for a screen.** Ask GitHub Copilot for a design, for example: Design a three-tier pricing section in Brilliant.
3. **Pair the canvas.** When Copilot's agent calls Brilliant, brilliant.design opens in Chrome, Edge or Firefox and pairs with it. Allow the one-time local network prompt. If the Brilliant desktop app is running, Copilot's agent uses it instead.

## What can GitHub Copilot do on the canvas?

Everything the editor does.

- **Create** Native editable elements, from HTML and CSS or from Blueprint.
- **Edit** Any canvas: read structure, change properties, reparent, rename, delete, reorder.
- **Search** Across canvases by name, text, type, color or component.
- **Run commands** Over a hundred canvas commands, each with full undo.
- **Export** PNG, JPEG, WebP, SVG, PDF, HTML, React, MP4, MOV and Replay.
- **[Generate](https://brilliant.design/docs/ai/what-the-ai-can-do)** Raster images and vectors, and vectorize a raster into editable paths.

## What can I do by hand afterwards?

Everything Copilot's agent makes is a real, editable element, never a flattened image.

- **Drag** Move and resize frames, and change their auto layout.
- **Retype** Edit any text where it sits.
- **Restyle** Change colors, fills, strokes, effects and components.
- **Undo** Each command can be undone, so a wrong move is one keystroke back.
- **Hand it back** Copilot's agent reads your changes and keeps going from them.

## What does it need?

Node 20 or later, for the helper the command installs. A browser canvas in Chrome, Edge or Firefox. Safari cannot run the browser canvas.

Or the Brilliant desktop app for macOS or Windows, which serves MCP on your machine while it is open. The helper and the desktop app serve MCP on your own machine, so your design does not reach Brilliant over them, except an image generated on Brilliant AI and background removal.

[Connect an agent, every client](https://brilliant.design/docs/agents/overview)

## What does it cost?

Brilliant has a Free plan, and canvas work by your own agents over MCP is unlimited on every plan, Free included. An image an agent generates on Brilliant AI uses your Brilliant AI usage.

[See pricing](https://brilliant.design/pricing)

## Questions

### How do I get GitHub Copilot to design a UI?

Connect Brilliant with one command, `npx -y @brilliant-hq/mcp install vscode`, then ask GitHub Copilot for the screen you want. Copilot's agent lays it out on a real canvas as frames, auto layout, components and text, and you can edit every part of it by hand.

### How do I make a mockup with GitHub Copilot?

Connect Brilliant, then ask for the mockup, for example: Mock up a mobile login screen in Brilliant. Copilot's agent builds it as editable layers. Move and restyle it by hand, ask for the next version, and export PNG, SVG, PDF, HTML or React when it is ready.

### Can GitHub Copilot make a poster?

Yes. With Brilliant connected, ask GitHub Copilot for a poster at the size you need. Copilot's agent lays out the text, shapes and images as editable elements, and you export it as PNG, PDF or SVG.

### How do I add Brilliant to GitHub Copilot by hand?

Add a brilliant entry under `mcpServers` in `~/.copilot/mcp-config.json`, the Copilot user file, or in `.mcp.json` at the workspace root: `"brilliant": { "type": "stdio", "command": "npx", "args": ["-y", "@brilliant-hq/mcp"], "tools": ["*"] }`. Or run MCP: Add Server from the Command Palette and choose Copilot Global. Then ask Copilot for a design and check that the tool call came from brilliant. The one-line command writes the older file instead, the VS Code profile mcp.json with the entry under `servers`: VS Code's reference calls that the Local harness format, deprecated and due for removal in February 2027, so new setups belong in the Copilot file. Format read at VS Code's MCP configuration reference on November 20, 2026.

### Do I need an API key?

No. The helper needs Node 20 or later and no API key.

### Is it free?

Yes. Brilliant has a Free plan, and canvas work by your own agents over MCP is unlimited on every plan. An image generated on Brilliant AI uses your Brilliant AI usage.

## Related

- [GitHub Copilot setup guide](https://brilliant.design/docs/agents/copilot)
- [MCP design tool](https://brilliant.design/mcp)
- [MCP servers for design work in 2026](https://brilliant.design/blog/mcp-servers-for-design)
- [Design canvases for coding agents, compared](https://brilliant.design/blog/design-canvases-for-coding-agents)
- [The Blueprint language](https://brilliant.design/docs/developers/blueprint)
