Ego Browser: give the agent a workspace, then check its work

Описание и применение

Ego Browser позволяет агенту работать в изолированном пространстве (Space), сохраняя контекст пользователя чистым. Агент выполняет задачи, создаёт снимки и проверяет результаты, а затем документирует источники и неопределённости.

Задача

Separating an agent’s browsing context from the human’s to avoid clutter and unintended interactions.

Результат

Agents can perform focused, read‑only or controlled tasks in isolated spaces, producing verifiable tables and screenshots while keeping the human’s browsing context clean.

Рабочий процесс

• Create a task space with a persistent ID
• Assign page labels within the space
• Navigate the agent to target URLs
• Perform actions (click, fill, screenshot)
• Capture structured snapshots and screenshots
• Verify post-action state (e.g., navigation, form values, persistence)
• Document source URLs and uncertainties in a table
• Close or reuse the space for subsequent rounds

Доказательства

An agent gets its own browser space while the human keeps their own browsing context.

Источник

The agent can use structured page snapshots, navigate, click, fill controls and capture screenshots.

Источник

After navigation, check the destination. After filling a field, check its value. After an authorized save, reload and check that the result persisted.

Источник