Ai Bridge

AI Browser Bridge

A minimal, open-source bridge that lets an **AI coding agent** (Claude Code, or any local tool that can run a shell command) drive **your real, logged-in browser** — safely, on localhost, with no cloud in the loop.

Born from a practical problem: an AI agent working in a terminal needed to download email attachments, fill review forms, and drive web editors on sites the user was logged into. OS-level input synthesis is platform-bound, steals window focus, and its synthetic events are rejected by CSP-strict web apps and rich-text editors. This bridge fixes all of that with three small pieces:

AI agent / your scripts          bridge server               Chrome extension
   `bridge ` CLI  ⇄  ws://127.0.0.1:8765 (token)  ⇄  service worker (MV3)
                                                             │ chrome.tabs / scripting
                                                             │ chrome.debugger  → trusted input, CSP-proof eval
                                                             │ chrome.downloads → authenticated downloads

Why the `chrome.debugger` route matters

• Trusted input. Input.dispatchMouseEvent / Input.insertText produce events with isTrusted: true — rich editors (contenteditable, ProseMirror- and Slate-class apps) accept them where synthetic DOM events are ignored.
• CSP-proof eval. Runtime.evaluate works on pages whose Content-Security-Policy blocks injected eval.
• No focus stealing. Everything runs in background tabs (newTab opens with active:false); your work is never interrupted.
• Your sessions, no re-login. The extension lives in your normal profile, so webmail, dashboards, internal tools — anything you're logged into — just work. This also sidesteps Chrome 136+'s restriction on --remote-debugging-port with the default profile.
• Cross-platform. macOS, Windows, Linux — no OS-level input scripting or accessibility APIs required.

Stealth by d