What is ego (lite)

What ego (lite) does and how it differs from other ways to give an AI agent a browser.

ego (lite) is a Chromium browser that you and your AI agent can use together on Windows. You browse as you would in Chrome. Your agent uses the same browser to research, fill forms, collect information from websites, or test a web app.

The agent connects through the ego-browser skill. Codex, Claude Code, and other agents that can run local commands can use it. Each task runs in a Space, so you can keep using your tabs while the agent works.

How ego (lite) works

Each of these features has its own page under Features.

Space

A Space holds the tabs for an agent's task inside ego (lite). It is part of the browser you are already using, not a separate window, cloud session, or headless browser. The agent opens and works with pages in its Space without switching your active tab or taking keyboard focus.

Spaces can use the logins you have in ego (lite). You can open a Space to watch a task, take over when a site needs your input, or review its pages afterward. Several tasks can run in separate Spaces at once, subject to the memory and CPU available on your computer.

Snapshot

A Snapshot is a compact text view of a page for the agent. It includes the page's content and gives controls short references, such as @3 for a field or @5 for a button. The agent reads this instead of sending an entire page of raw HTML to its model.

For example, a login page might look like this:

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 can use fillInput('@3', '...') and click('@5') to act on those controls. After the page changes, it takes a fresh Snapshot because the references may have changed too. Snapshots also help the agent read complex pages with embedded content, shadow DOM, or dynamically added controls.

ego-browser

ego-browser provides the tools an agent uses to open pages, read Snapshots, click, fill fields, and report results. The agent can put several steps in one JavaScript flow. You describe the task in your agent's chat; you do not need to run the code yourself.

await useOrCreateTaskSpace('read product news')
await openOrReuseTab('https://example.com', { wait: true })
cliLog(await snapshotText())

The agent can read the page, decide what to do next, and return details you can check. Running related steps together can reduce the back-and-forth between the agent and the browser.

Skills

The ego-browser skill teaches a compatible agent when and how to use these browser tools. After setup, check whether it appears in your agent's skills. Select it before giving the agent a browser task.

How it compares with other browser tools

Different tools suit different jobs. The main difference is where the browser runs and which tabs and logins the agent can use.

Browser extensions connect an agent to a browser window. In ego (lite), the agent uses its own Space without pairing an extension or taking over your active tabs.
Cloud browsers run on a provider's servers. ego (lite) runs on your computer and uses your local browsing data.
Built-in agent browsers belong to one AI app. ego (lite) is a full Chromium browser that more than one compatible agent can use.
Automation frameworks and agent CLIs often control a browser with a separate profile. ego (lite) lets the agent use your signed-in browser and combine several actions in one call.

Compared with browser extensions

Extensions are useful when you want help on the page you are already viewing. For a longer task, they may need to use that same window or keep a connection between the agent and the extension.

With ego (lite), the agent's pages stay in a Space. Your current tabs and keyboard focus remain yours. There is no extension to pair with the agent, and you can open the Space whenever you want to see its progress.

Compared with cloud browsers

A cloud browser runs on someone else's servers. That is useful for jobs that must continue while your computer is off or for running many browser instances on server infrastructure.

ego (lite) runs on your own computer. The agent works in that local browser instead of a hosted session. A site can still ask for a CAPTCHA or another security check, and you can handle it in the agent's Space. There is no charge for browser time in ego (lite).

Compared with built-in agent browsers

A browser inside one AI app is convenient for quick tasks, but it may have a separate login state. You may need to sign in to the same sites again, and another agent cannot use that browser.

ego (lite) is a browser you can use directly, with your own tabs and the sites you sign in to. Codex, Claude Code, and other compatible agents can work in separate Spaces while you keep browsing. Switching agents does not require switching browsers.

Compared with automation frameworks and agent CLIs

Frameworks such as Playwright, Puppeteer, and Selenium work well for repeatable test scripts. Agent CLIs give an AI agent browser commands, often against a fresh browser profile. Tasks on signed-in sites may need extra login setup.

With ego (lite), the agent can start from a login you already have in the browser. It can open a page, read it, and perform several actions in one ego-browser call. You can watch its Space and take over for a QR code, verification prompt, payment, or another step that needs your decision.

Privacy

ego (lite) runs on your computer instead of in a hosted browser. If setup offers an option to import data from another browser, review the profile and data shown before confirming.

Your agent may send the page content it reads to the model provider you use with that agent. The provider's data policy still applies. Review what the agent can access, and ask it to pause before publishing, deleting, paying, or submitting changes.

Platforms and price

ego (lite) is available for Windows. It is free to download and use, with no subscription, per-task charge, or browser-time fee. You use your own agent and model provider, which may have their own costs.

Start with Quick start to install ego (lite) and try a browser task.