Connecting an Agent
Caja exposes an MCP (Model Context Protocol) server that AI agents can connect to.
Supported Clients
Caja can be configured with any MCP-compatible client:
- Claude Code
- Claude Desktop
- Cursor
- VS Code
- Codex
Quick Setup
- Open Caja
- Click the plug icon in the title bar
- Select your client
- Click Install — Caja writes the config automatically
Manual Configuration
Add Caja to your client's MCP settings:
json
{
"mcpServers": {
"caja": {
"command": "caja",
"args": ["--mcp"]
}
}
}Config File Locations
| Client | Path |
|---|---|
| Claude Code | ~/.claude/ (project or global settings) |
| Claude Desktop | ~/Library/Application Support/Claude/claude_desktop_config.json |
| Cursor | .cursor/mcp.json in your project |
| VS Code | .vscode/mcp.json in your project |
Verifying the Connection
When an agent connects, the plug icon in Caja's title bar changes to indicate an active connection. A spinning loader appears when the agent is executing a tool.
What Agents Can Do
Once connected, agents can:
- Read the frame tree and take screenshots
- Add, update, and remove elements
- Manage pages and components
- Apply Tailwind classes
- Execute multi-step operations in batch
See the Tools Reference for the complete API.