Getting started

Three steps: register the DreamUI adapter in your harness, sign in, and run a generation.

Prerequisites

A DreamUI account at dreamui.com.
An MCP-compatible harness running on a machine you trust.
The adapter configuration from the DreamUI ZIP guide.

Add DreamUI as a stdio MCP server using the generated configuration from the ZIP guide. It points at https://dreamui.com and keeps process paths on the host. Where that configuration goes depends on the harness:

Where each harness reads MCP configuration
Harness	Configuration
Claude Code	`claude mcp add, or a project .mcp.json`
Codex	`[mcp_servers] in ~/.codex/config.toml`
Cursor	`~/.cursor/mcp.json or .cursor/mcp.json`
Windsurf	`~/.codeium/windsurf/mcp_config.json`
VS Code	`.vscode/mcp.json`
Meta AI Muse	`See docs/mcp/muse-guide.md`

Any other MCP client works the same way: register a stdio server with the generated command.

Sign in
Ask your agent to sign in to DreamUI. It calls uicopilot_auth, opens a login link, and stores the credential locally once you approve. Details in Device login.

Run a generation

Ask your agent to build something. Under the hood it makes a call like this:

MCP · tools/callFirst generation with uicopilot_agent_run

{
  "name": "uicopilot_agent_run",
  "arguments": {
    "agent": "code",
    "prompt": "Create a responsive product landing page with a pricing section.",
    "localDir": "/absolute/path/on/the-mcp-host/project",
    "writeFiles": false,
    "timeoutSec": 180
  }
}

Uses credits. writeFiles: false returns output without writing files to localDir.

Next steps

Maintained guides

Maintained source guides
File	Covers
`README.md`	Product overview and supported user workflows
`docs/CLI.md`	CLI behavior and authentication
`docs/MCP.md`	MCP adapter setup and the admin MCP route
`docs/mcp/muse-guide.md`	Connecting DreamUI to Meta AI Muse

Prerequisites

Setup

Next steps

Maintained guides