Agent Browsing Moved to the Cloud: Sessions, Locations, Handover, and the Per-Minute Bill

For agents to work on websites, Octomind, the company behind Octos, has moved the browser to the cloud. Each Octo now has its own cloud browser, with its own profile, location, and the ability to view the keyboard in real-time. When an Octo signs in, the session begins and lives in the cloud, separate from the agent's personal sessions.

Agents can choose the location for the cloud browser, such as the United States, United Kingdom, Germany, France, Singapore, or no fixed address. The location setting sticks, ensuring consistent results for tasks that depend on location. The handover feature allows agents to take control of the browser for a few seconds to handle tasks that require human input, rather than writing workarounds for specific cases.

Octos browse at a rate of $0.02 per start and $0.015 per minute, with the last minute rounded up when the browser closes. This pricing model allows agents to batch multiple pages into a single session, reducing costs compared to opening a browser for each step. The conversation header now displays the current and total spending, with detailed spending information available upon request.

Agents can now submit forms to provide the additional details required by certain tools, without needing to interact with the Octo. The conversation flow has been streamlined, eliminating the need for unnecessary prompts and questions. Paid plans now support more model calls in parallel, increasing efficiency. The latest version of Octomind is 0.67.0, running on a more efficient model call system and allowing multiple files to be read simultaneously.