Perplexity has introduced a new feature in its agent Comet — the model now has the ability to take screenshots while operating in the browser. The screenshots instantly appear in the task list, allowing everyone to verify what the model “saw” during its actions. This approach opens up a new level of transparency and significantly simplifies error detection in complex automated processes.
The update will be a real find for those who trust the agent to browse or collect data from websites. Now users receive not only text confirmations but can also visually assess the context — this is especially important when working with rich interfaces, graphic reports, or unstructured content. Additionally, this feature allows the model to provide short descriptions or summaries based on visual information.
Simultaneously, Perplexity is preparing integration with Outlook for desktop environments, primarily for Windows. Previously, the model had already learned to work with Gmail and Google Calendar, and Outlook will open the door to automation for corporate users and office tasks. Considering that the Comet browser is expected to appear on Windows soon, this combination seems like a logical step in the company’s strategy.
The version of Comet for Android is still set to launch in the fall. This will expand the model’s capabilities in mobile automation and help cover even more usage scenarios. Thus, Perplexity is gradually transforming Comet into a universal agent capable of working with emails, calendars, and web interfaces, providing a complete cycle of task execution and control across different devices.
