Tab Agent — Agentic AI for your Browser

Tab Agent is a Chrome extension that runs an AI agent directly in your browser — not just chatting about a page, but actually reading it, clicking, typing, navigating, and hopping between tabs until a task is done. Bring your own Anthropic or OpenAI-compatible API key. There's no backend, no middleman server, and no account to create. Your key talks directly to your chosen provider from your browser, and your conversations never pass through any server the developer operates. What it can do: - Complete multi-step browsing tasks: click, type, navigate, and read pages on your behalf, showing every step along the way rather than acting as a black box. - Research and compare across multiple tabs in parallel, instead of reading one page at a time. - Run scheduled checks in the background — watch a price, a job posting, or a status page, and get notified only when it matters. - Look at what's actually on a page: screenshots, images, and generated images, not just text. - Read long attached documents (PDFs and more) in full, without silently truncating content near the end. - Save custom @agent presets — a name, a target site, and instructions — for workflows you run over and over. Full source is public on GitHub: https://github.com/VeeraChandu/tab-agent