Agtweb

Let local software agents inspect and automate web pages in your browser.

Agtweb lets software agents running on your computer inspect and automate web pages in your existing Chrome, Chromium, or Firefox session. This is its single purpose. A compatible local agent can: - navigate and manage tabs, and inspect page frames; - read page content, interactive elements, and form fields, with structured Markdown available as one inspection format; - click, scroll, type, fill forms, and upload or drop files; - capture the visible tab and inspect recent download metadata; and - run bounded sequential or parallel page workflows and web searches. Actions run in the browser session you already use, including pages where you are signed in. Agtweb accepts only named actions implemented in the submitted package. It does not execute client-supplied JavaScript, contain an AI model, choose tasks, or act without a local request. Web pages cannot issue automation commands. Installing Agtweb grants the browser permissions shown by Chrome, Chromium, or Firefox. Agtweb adds no per-site approval prompts. The local bridge accepts only clients that present the random token created by its installer. Agent automation uses the separately installed `agtweb-bridge` component. It connects through the browser's native-messaging API; the provided installer exposes a token-protected WebSocket on `127.0.0.1` for local clients. The extension has no developer backend, analytics, or telemetry. Requested browsing and interaction still contact destination sites as ordinary browser use. The toolbar icon only reports bridge connection and activity; it opens no popup and starts no action. The companion `agtweb-bridge` is installed separately, contact support@emee.ai for details.