Source: devblogs.microsoft.com/foundry/browser-automation-tool-ignite
The Browser Automation tool in Foundry Agent Service, which runs on Azure Playwright Workspaces, is now generally available. Each Foundry agent gets scalable browsers with isolated sessions. Its MCP tools include step-by-step controls for clicking, typing, form filling and screenshots, plus a run-code tool that executes one Playwright script for a batch of actions.
Human oversight and observability features are generally available. Live View shows the agent's browser in headed mode, and Take control lets a person complete a sign-in or approve an action before handing control back. Session recordings are opt-in and stored in the customer's own storage account, and activity logs and billing observability are also available.
New preview features include VNet injection for reaching private applications, a bring-your-own browser proxy with domain-based routing and credentials held in Azure Key Vault, and secret management that keeps credentials away from the model. Profile reuse keeps cookies and sign-in state between sessions, and Web Bot Auth lets agents sign their HTTP requests, with registration through self-service or selected bot management providers.
Why it matters if you build agents
Teams building browser agents can now automate internal portals and private applications while keeping traffic inside their own network controls and keeping credentials out of model context. Batched run-code scripts can reduce the number of calls in repetitive flows, but they can run any code, so Microsoft recommends enabling them only for reviewed scenarios. Preview features such as profile reuse and CAPTCHA handling are subject to change, so check their documentation before using them in production.
Get stories like this every weekday
One short email with the day's agent tooling news.