Egg, the browser

Egg is a Chromium-based browser with the platform built in. This track covers how it behaves as a browser, and how to build for and automate it.

Egg runs as two coordinating processes on the user’s machine: a browser shell that owns tabs, chrome, and the UI, and a long-running daemon that hosts the agent, polls feeds, runs scheduled work, and serves a local HTTP API. The split lets the agent keep working when no window is open, and keeps long-running services out of a process that exits when the UI closes.

The web engine is the platform’s native one: WebView2 (Chromium + Blink) on Windows, WKWebView (WebKit) on macOS. No embedded second engine, no pinned binary. Page-side compatibility (Service Workers, WebAssembly, PWAs, Chrome extensions) follows what each engine supports.

Browsing stays on the device, and engine-level telemetry is disabled wherever the API allows. Multi-device sync rides an encrypted relay that only ever sees ciphertext. A URL-pattern firewall gates what an agent may do when it reaches into the browser, with conservative defaults on sensitive categories.

Where to start

Architecture is the full picture of how Egg behaves as a browser. Browser Automation covers driving it over CDP, and Headless & CLI covers the command line. The browser’s HTTP surface lives alongside the gateway’s in the HTTP API.