DOMino

DOMino is a voice-driven browser agent built as a Chrome Extension that turns spoken commands into real browser actions. Hold a key, speak naturally, and DOMino captures your screen, reads the page as a numbered map of every interactive element, and autonomously clicks, types, navigates, and completes whole tasks through a tight observe, reason, act loop. It works on virtually any website because it targets elements by index rather than brittle selectors, making it robust across layouts and dynamic pages.

The system runs entirely on your local machine with a FastAPI backend handling the agent loop, transcription, and SSE streaming. Reasoning is powered by Gemini via Vertex AI with multimodal vision and language understanding, and the architecture is fully pluggable supporting AWS Bedrock, OpenRouter, Groq, and Claude as alternative providers. Speech is handled by ElevenLabs for both transcription and voice readback, with Groq Whisper and Deepgram as fallbacks. The extension is built with TypeScript, React, and Webpack targeting Chrome Manifest V3. Firecrawl converts pages to clean markdown for richer context, per-tab conversational memory handles multi-turn follow-ups, and scoped batching reduces model calls without sacrificing accuracy. No data ever leaves your machine.