Source: devblogs.microsoft.com/foundry/whats-new-in-microsoft-foundry-october-2026
On October 28, Microsoft announced the Microsoft Copilot harness in public preview in Foundry Agent Service. It gives prompt agents a managed loop and sandbox for tasks that involve reading files, running code, using supported tools and applying reusable skills. Developers still choose a supported model, and the harness must be selected when an agent is created.
Long-running resilience for hosted agents reached general availability on the same date. After an interruption, Foundry can restart work with its original input, so applications need to save intermediate results, completed steps and identifiers to avoid repeating external actions. Browser Automation became generally available on October 19 and runs through a separately provisioned Playwright Workspace.
GPT-6.1 Terra became generally available on October 26 and accepts text and images through the Responses API. The Decisions API reached general availability on October 22 and returns typed answers to predicate, choice and rubric questions. MAI Image 2.6 is in preview in six regions.
Why it matters if you build agents
Agent builders can use a managed loop instead of maintaining their own, but checkpoints and duplicate-action checks still belong in the application. Browser workflows should use least-privileged identities and require human approval before consequential actions. Routing decisions can now return typed answers, which should be tested against labeled examples before they trigger actions.
Get stories like this every weekday
One short email with the day's agent tooling news.