What is ego lite
What ego lite is, who it's for, and how it differs from existing browser automation tools.
ego lite is a browser built for you and your AI agents to work together. Your agents run tasks in parallel, each in its own Space, share Chrome logins with you, and stay out of the tabs you're using. The same tasks finish faster, on fewer tokens.
Any agent CLI can plug into ego lite through the ego-browser skill: Claude Code, OpenAI Codex, Cursor, Continue, Gemini CLI, Hermes Agent, OpenClaw, Opencode, or your own custom agent.
Core features
A short overview here — each feature has its own page under "Features".
Space
A Space is an isolated workspace inside ego lite. It's not a new window, not another Chrome profile, and not headless mode. It's a partition inside the same browser where the agent does its work without touching the tabs you're using.
You and the agent share one browser. The agent runs in its Space, you keep working in your tabs, and neither side steals focus from the other. You can switch into a Space anytime to see what the agent is doing, take over, or stop the task.
A Space can hold many tasks in parallel; how many depends on your machine, with one Space per agent or per task. Work that used to queue up in a single browser now runs side by side: Claude Code enriches 10 leads across 10 Spaces, Codex scrapes 5 competitor sites in another 5, and you keep working in your own tabs. None of these tasks collide on tabs, snapshots, refs, or browser state, and none of them touch yours.
→ Space
Snapshot
ego lite ships with a deeply customized Chromium kernel that produces high-quality page snapshots — the view text models use to "see" and "act on" a webpage. A typical page reaches the agent as a compact accessibility tree, with a short ref on every element. A whole page usually fits in 200–400 tokens, far less than raw HTML.
Calling snapshotText() on a login page returns something like:
Page: Example - Log in
URL: https://example.com/login
@1 [heading] "Log in"
@2 [form]
@3 [input type="email"] placeholder="Email"
@4 [input type="password"] placeholder="Password"
@5 [button type="submit"] "Continue"
@6 [link] "Forgot password?"
The agent immediately knows @3 is the email field and @5 is the submit button, and acts on the refs: fill('@3', '...'), click('@5'). No CSS-selector guessing, and class-name rotations don't break anything.
Snapshots are produced inside the Chromium engine rather than by a JavaScript shim layered on stock Chrome. That means cases shim-based tools quietly miss — deeply nested iframes, shadow DOM, and third-party SDK components like Stripe checkout, Salesforce embeds, Intercom panels, or React portals mounted dynamically — work reliably here.
→ Snapshot
ego-browser
This is how the AI agent drives the browser. Navigation, observation, action, and output all live inside a single Node.js heredoc that runs in one shot:
ego-browser nodejs <<'EOF'
await useOrCreateTaskSpace('follow-ego-agent')
await openOrReuseTab('https://x.com/ego_agent', { wait: true })
await snapshotText()
await click('@92') // @92 is the Follow button from the snapshot
cliLog('Done.')
EOF
One heredoc, one round trip — instead of "run a command, read the output, run another, read more output." In our internal benchmark across four complex browser tasks (against Vercel's agent-browser), ego lite finished each task up to 245% faster and used noticeably fewer tokens. The harder the task, the wider the gap.
Experience accumulation (coming soon)
Every task that finishes successfully gets distilled into a reusable artifact — a tool or a workflow — filed by domain. The next time the agent runs a task on the same site, it loads the validated solution and skips the trial-and-error. Internal tests show the second run of a complex task can be up to 5× faster than the first, on top of the gains ego-browser already delivers in a single run.
→ Skills
Privacy
Your browsing data stays on your machine. Passwords, cookies, history, bookmarks, extensions, and Profile imported from Chrome never leave your device. Page contents are read and acted on by your own agent. ego lite handles the browser and the bridge; the thinking and reading are your agent's job.
The only thing ego lite records at install time is whether you chose to import data from Chrome.
No account to register, no email to collect, no cloud session to sign into.
Platforms and price
macOS today. Windows and Linux are on the roadmap.
Free to download. No subscription, no per-task billing, no cloud session metering. Bring your own Agent CLI and your own model API keys.
Further reading
Why we built ego lite and how we designed it, in this blog post: A Browser for You and Your AI Agents to Work in Parallel.