Microsoft opens its Foundry browser tool for AI agents to all users

Microsoft tells that all users of Foundry can use its browser tool for agents. The tool gives an agent a browser that Microsoft operates. The agent can open sites, read information, fill in forms and do tasks of many steps from written instructions. The tool works for websites and for apps in a network of a company. Microsoft gave no test results or use numbers.

How it works

The agent looks at the page, selects the next step, does it and checks the result before it continues. The browser operates in the cloud. Teams install no browser. Teams must make a different Azure resource, Playwright Workspaces, for the browser. Access to a network in the company is added at that time.

The controls

Teams give each agent only the access for its task, with Microsoft Entra identities and role based access control. Tools that change a page can stay off at first. Live View shows the browser as it works. Take control lets a person control the agent.

What is not known

Microsoft gives no numbers for speed or for how frequently agents do the task correctly. It gives no result for a page that changes.

Insight: Companies sell control of agents, not only agents

Microsoft supplies this browser with access limits, approval steps and a person who can control the agent. Google tells that a group that uses its Gemini Agent can set identity, audit and spending rules at one point. This reads as a toll gate: the company that operates the browser of the agent also holds the control point that other teams must use. Teams that select where the agent works also select who holds that control.