gemini-skill

Why we included this project

Agents that can only talk are half useful, and this project gives them Gemini's image generation and chat through the browser, without paying for a separate API tier. It runs a standard MCP server plus a skill, so any MCP-capable client can call tools that drive a real logged-in browser session over CDP. The browser daemon handles the tedious parts, like keeping the session alive, evading automation detection, downloading full-resolution images, and stripping watermarks. That makes it a drop-in way to add Gemini drawing and image extraction to an existing agent setup, and the OpenAI-compatible Atlas Cloud provider offers a pure API route if you want to compare browser automation against it. Just know it needs a locally installed browser and a logged-in Google account, so it suits environments where that setup is acceptable.