Copilot Studio adds Anthropic’s Claude to its agent lineup

For the past two years, building an AI agent inside Microsoft's walled garden basically meant one thing: you were running on OpenAI. That exclusive bond just loosened up in a major way.

Microsoft announced that it is bringing Anthropic's Claude models directly into Microsoft Copilot Studio and Microsoft 365 Copilot, giving enterprise users a genuine choice beyond OpenAI for the first time.

Claude enters the enterprise cockpit

The integration lands across two primary surfaces: Copilot Studio and the Researcher agent in M365 Copilot.

Inside Copilot Studio, builders can now pick Claude Sonnet 4 or Claude Opus 4.1 from a simple dropdown menu alongside OpenAI models. You can assign Claude to handle primary orchestration, chat, or deep reasoning workflows.

Even better, multi-agent systems can now mix and match. You can build an HR onboarding assistant where one model answers casual employee questions, while another handles tricky compliance verification and policy interpretation.

Meanwhile, the Researcher agent in M365 Copilot lets users select Claude Opus 4.1 or OpenAI's deep reasoning models to crunch multistep research across emails, chats, documents, and web data.

The models are rolling out now to early release environments worldwide, heading to public preview within two weeks, and slated for full production readiness by the end of the year.

Admin controls and the safety net

Microsoft is keeping corporate admins firmly in the driver's seat.

Anthropic access requires an explicit opt-in inside the Microsoft 365 Admin Center. Once enabled there, admins can manage tenant-level maker access inside the Power Platform Admin Center.

There is also a built-in safety net: if an admin disables Anthropic models down the road, existing agents will automatically fall back to OpenAI GPT-4o without breaking workflows.

Microsoft noted that Anthropic models are hosted outside Microsoft-managed environments and remain subject to Anthropic's Terms of Service.

Why it matters

For enterprise teams, this removes friction where it hurts most. Many developers and knowledge workers already prefer Claude for complex reasoning and nuanced writing, but bringing external API keys into strict enterprise workflows was a compliance nightmare. Now it is just another native toggle.

For Microsoft, this marks a pragmatic pivot away from OpenAI exclusivity toward becoming an agnostic orchestration layer. When clients want the best tool for a specific job, denying them Anthropic only risks losing them to competing platforms.

We think this is a smart defensive move that keeps customers locked into Copilot Studio as the primary control tower. The catch? Because these models run outside Microsoft-managed infrastructure under Anthropic terms, risk-averse legal teams in heavily regulated sectors might still hesitate before flipping the master switch.