AI Agent Browser 'ego lite' Released

Existing browser automation tools have revealed several problems when performing parallel tasks. For example, when using tools like Playwright, mouse cursor control is often hijacked, or browsers are launched indiscriminately and not properly shut down. These issues significantly hinder work efficiency, especially when multiple AI agents perform tasks simultaneously, as they contend for ownership and cause conflicts.

Ego lite addresses the interference problem between AI agents by providing a completely independent workspace, or 'Space,' for each AI. Within a single browser executable, each AI performs tasks only in its assigned space, so a running AI will never take focus or intervene in another AI's work. This allows various AI agents, such as Claude or Codex, to perform multiple tasks simultaneously without conflict.

The installation and AI integration of ego lite are very simple. After downloading and running the macOS installation file from the official website, an 'Ego Browser' skill is automatically created within the AI agent environment (such as Code or Claude) that the user has already installed. Subsequently, users can call 'Ego Browser' via a slash (/) command in the AI agent or check the immediate connection status by making a request like "Check Ego Browser access."

JavaScript-based browser automation is essential when an API for a specific service is not provided or when complex, multi-step procedures need to be automated. For example, to manipulate filters on a logged-in dashboard, understand usability issues of a new service, or have an AI directly execute multi-step reservation processes, functionality that allows actual manipulation within the browser is required. Ego lite, leveraging its Chrome-based engine, meets these requirements.

Ego lite is a standalone browser developed directly using Chromium, the engine of the Chrome browser. Therefore, it is not merely a Chrome extension but provides a complete browser environment in itself. When users launch ego lite, a very familiar screen appears, and when connecting to websites like google.com, pages load identically to a regular Chrome browser. This signifies independent operation, distinct from the existing Chrome environment.

Ego lite supports efficient execution of repetitive tasks by maintaining user login sessions, cookies, extensions, and browser environment settings. Instead of creating a new session every time, ego lite functions like a regular Chrome browser where information persists when navigating between pages. This allows AI agents to automate tasks such as capturing or data extraction in a pre-logged-in state, eliminating the hassle of logging in each time and speeding up work.

One of ego lite's core features is 'Space'. A number displayed in the upper right corner of the browser indicates the connection status of the currently running AI agent. By clicking the 'Takeover' button within this space, the user can directly take control of the browser. Conversely, selecting 'Return Agent' allows the AI agent to reaccess the browser and continue its task. This feature enables free transfer of control between AI and user, and each AI agent can perform tasks independently in its assigned dedicated workspace, enabling efficient automation without mutual interference.

Ego lite's browser automation feature can be useful for testing web services where API integration is difficult or inefficient. For example, an AI agent can be instructed to automatically perform the process from selecting a flight to entering test passenger information and up to the pre-payment stage. This makes it possible to test and verify complex reservation processes in an environment similar to actual user experience.

AI agents can be used to freely explore and debug web services without predefined scripts. An AI can be instructed to explore a service as if it were a first-time user with a test account, checking all major functionalities across desktop and mobile sizes. If an issue is found, it can attempt various experiments to identify the cause and generate a detailed report with screenshots for the user.

Ego Browser fundamentally solves problems arising from parallel processing by utilizing independent spaces. Even when multiple browser instances are running simultaneously, each task proceeds within its isolated space, preventing mutual interference such as mouse cursor control conflicts. Even if several additional browsers run alongside those already managed, they do not affect each automation thread and can perform tasks stably.

Ego lite provides a browser control transfer feature that allows human intervention during AI automation tasks. Pressing the 'Takeover' button allows the user to immediately take control of the browser to perform necessary actions, and control can be returned to the AI agent at any time. This enables flexible workflows where users can directly intervene in sensitive areas requiring login or other authentication, and then hand automation back to the AI.

Ego lite enables the extraction and analysis of analytical data from web services that were previously inaccessible via API. For example, various charts provided in YouTube Studio are often difficult to access directly through an API. However, using ego lite, this chart data can be collected to specifically analyze where and why user churn occurs, making it highly useful for data-driven decision-making.

Ego lite can maximize the efficiency of end-to-end testing and significantly reduce time through its parallel testing feature. Pages to be tested can be listed, divided into multiple groups, and each group assigned a browser automation prompt to run simultaneously. For instance, running parallel tests divided into six groups can reduce testing time by approximately 1/6 compared to traditional methods, greatly accelerating the overall development and verification process.

Ego lite offers distinct advantages that differentiate it from specialized testing tools like Playwright. While existing tools may be more suitable for repetitive verification tasks, ego lite allows multiple AI agents to perform tasks simultaneously without interference, maintaining logged-in sessions in a browser with actual usage history. This is ego lite's biggest advantage in automating complex tasks by allowing various AI agents to cooperate organically in a constantly changing web environment.