Gemini CLI

Gemini CLI connects to ReUI in two quick steps: point it at the hosted MCP server, then sign in and start building UI by chat. The endpoint is https://mcp.reui.io, a remote server over Streamable HTTP, so there is nothing to run locally.

Connect ReUI

{
  "mcpServers": {
    "reui": {
      "httpUrl": "https://mcp.reui.io"
    }
  }
}

{
  "mcpServers": {
    "reui": {
      "httpUrl": "https://mcp.reui.io",
      "headers": {
        "Authorization": "Bearer $REUI_LICENSE_KEY"
      }
    }
  }
}

Install the ReUI skill

Run the one-line installer in your project so Gemini CLI follows ReUI's build workflow:

curl -fsSL https://mcp.reui.io/install | node -

For Gemini CLI it writes the skill to .agents/skills/reui, where Gemini CLI looks for workspace skills, and the reui server entry to .gemini/settings.json, so the first step above is done for this project too. If the project already has a GEMINI.md, it also adds ReUI's short design-rules block there, between <!-- BEGIN:reui-agent-rules --> and <!-- END:reui-agent-rules --> markers, and changes nothing outside them; it never creates that file. The same run writes the files of the other agents it supports, which do nothing if you do not use them; one of them, the reui entry in Codex's ~/.codex/config.toml, is in your home directory rather than the project.

Put REUI_AGENTS_MD=0 before node (curl -fsSL https://mcp.reui.io/install | REUI_AGENTS_MD=0 node -) to leave GEMINI.md and AGENTS.md alone, REUI_AGENTS_MD=remove to take the block back out, or REUI_DRY_RUN=1 to print every planned write without writing anything. REUI_GLOBAL=1 writes ~/.agents/skills/reui and ~/.gemini/settings.json instead.

Skip the install and the skill still reaches Gemini CLI through the MCP: ask it to call get_agent_skill for the core workflow and a list of topics, then ask for a topic (color, motion, layout, components) to read that reference in full. If a build drifts off-pattern, tell it to "follow the ReUI skill" to reload that workflow.

Build with ReUI

Describe the interface you want in plain language and Gemini CLI pulls the right ReUI components, examples and blocks from the registry, then wires them into your project. A few prompts to start with:

Use ReUI to scaffold an admin app - the app-shell with a collapsible sidebar and top bar, and a dashboard page with stat cards and a chart.

Add a data-grid of orders with sorting, column filters, pagination, and row selection, wired to my /api/orders endpoint.

Add a kanban board for support tickets with drag and drop, grouped by status, and a filters bar above it.

Unlock premium items

Free items (the 27 components and every c-* example) install with no license. To let Gemini CLI install premium blocks and icons too, grab a Pro or Ultimate license and add the key to your project:

REUI_LICENSE_KEY=your-license-key-here

{
  "registries": {
    "@reui": {
      "url": "https://reui.io/r/{style}/{name}.json",
      "headers": {
        "Authorization": "Bearer ${REUI_LICENSE_KEY}"
      }
    }
  }
}

Troubleshooting

• Tools missing or the server shows as disconnected - the connection did not finish. Check the config file for a typo in the URL, then restart Gemini CLI.
• 401 after signing in through the browser - your ReUI session expired. Sign in to reui again from the MCP panel.
• 401 with a personal token - the token has expired or been revoked. Signing in again will not help: create a new token at Account → MCP and update the header.
• Sign-in works but every call still 401s - an Authorization header on the server entry overrides the browser session. Remove it, or make sure it carries a valid token.