Clients

Gemini CLI

Gemini CLI / Google tooling — MCP config.

Prerequisite

Install the Screen Beaver Windows desktop app first. MCP only works after the software is installed and running — the agent talks to the studio over MCP; it does not replace the app.

Get from Microsoft Store, install it, then leave Screen Beaver open while you use your agent.

MCP CLI

After the desktop app is installed, install the MCP CLI once (shared by every client):

terminalshell
npm i -g @screenbeaver/mcp

Full npm notes: npm install page.

Wire screenbeaver-mcp into Gemini’s MCP config the same way as other local stdio servers.

Config

Config location: Gemini CLI MCP config (or ~/.gemini MCP settings)

mcp_config.jsonjson
{  "mcpServers": {    "screenbeaver": {      "command": "screenbeaver-mcp"    }  }}

Verify

Confirm Screen Beaver is running on Windows, then From the Gemini CLI or IDE chat, confirm tools and run a Screen Beaver demo.

Prefer outcome prompts: “Capture onboarding, add captions, generate voiceover, prepare export.” The agent should call record_and_build or live_build_demo.