ego lite: Agent-Native Chromium Browser Launch

ego lite released an agent-native Chromium browser designed for both daily human use and parallel agent driving. Agents such as Claude Code, Codex, and Cursor operate in isolated Spaces without interfering with user tabs or login states. All data remains local, with zero configuration. - Each agent gets its own Space with isolated tabs but inherits the user's login state, enabling simultaneous multi-agent tasks. - Space ownership types include agent, agentDelegatedToUser, and user; agents hand off to the user for 2FA/CAPTCHA and cannot auto-bypass. - Agents use in-page JavaScript functions (Playwright-style) instead of CLI round-trips, compressing observe-decide-act-wait-verify-extract into a single Bash call. - Complex tasks are up to 3.45× faster than Vercel agent-browser with significantly fewer tokens; simpler tasks also benefit with 2.5× speedups reported. - A kernel-level semantic snapshot reliably penetrates cross-origin iframes, Shadow DOM, and third-party SDK components (Stripe, Salesforce, Intercom, React portals) where JS shims fail. - One-click import of existing Chrome settings (bookmarks, passwords, extensions, cookies) makes ego lite usable as a default browser with no login friction. - First launch offers to migrate all Chrome data, preserving logins, cookies, extensions, and bookmarks. - Supports macOS, Linux, and Windows through Chromium foundation.
Key insights
Space isolation solves the conflict between human browsing and autonomous agent actions
ego lite introduces Spaces as the fundamental unit, giving each agent a separate environment with inherited login states. Up to 10 Claude Code Spaces for lead enrichment and 5 Codex Spaces for competitor scraping can run in parallel without mixing with the user's personal tabs. Authentication hurdles like 2FA and CAPTCHA trigger a handOff to the user, after which the agent takes over again, ensuring security without blocking automation.
Code-base execution model reduces latency and token consumption compared to CLI loops
Traditional browser automation sends a command, waits for output, then repeats, incurring a model round-trip each step. ego lite exposes Playwright-style functions as in-page JS, letting an agent construct a single heredoc that executes the entire observe-decide-act-wait-verify-extract cycle in one Bash call. This approach cuts complex task time by 3.45× versus Vercel agent-browser and uses fewer tokens, with the advantage growing as task complexity increases.
Kernel-level semantic snapshot bypasses common web automation failures in iframes and Shadow DOM
Instead of a JS shim on top of stock Chrome, ego lite's snapshot is built into a custom Chromium kernel, delivering faster, more reliable element references. It successfully handles cross-origin iframes, Shadow DOM, and embedded components from Stripe, Salesforce, Intercom, and React portals—scenarios where web fetch and shim-based approaches typically fail, causing agents to retry and burn tokens.
One-click Chrome migration eliminates login friction by inheriting existing authenticated sessions
ego lite prompts users to migrate Chrome data on first launch, importing bookmarks, passwords, extensions, cookies, and login states. Because all websites are already authenticated for the user, agents never encounter CAPTCHA, 2FA, or SSO barriers. This inheritance model eliminates the need to manage separate browser profiles or manually inject credentials.