ego-lite: A browser for writing alongside AI agents, 2.5x faster automation

Existing browser automation frameworks require agents to drive a separate browser, and login states are not properly passed, causing users to compete with agents for tabs. ego lite is a dedicated browser designed for users and AI agents to share their respective spaces within the same browser. It runs on macOS, and when Chrome data is migrated, agents inherit existing logins, cookies, extensions, and bookmarks as-is.

Agents directly invoke browser functions as JavaScript functions via the ego-browser skill. Instead of exchanging commands multiple times, multi-step tasks are composed into a single code snippet for execution, completing complex workflows up to 2.5x faster. Token consumption is significantly reduced, and success rates improve. Each agent works in an isolated Space, so it does not affect the user's tabs or mouse position.

Various agent CLIs such as Claude Code, Codex, and Cursor can access the browser through ego-browser. Core tools such as page snapshots, input, clicks, waits, navigation, and capture are provided, and complex structures like deeply nested iframes are handled reliably. Data is stored locally and released under the MIT license.

It is particularly useful when multiple agents perform parallel tasks simultaneously. For example, while Claude Code processes 10 leads in parallel Spaces, Codex can crawl 5 competitor sites. They do not conflict with each other, and the user can continue their own work. It is suitable for enhancing agent productivity in development environments with repetitive browser tasks.