Meet the new Copilot Studio: rebuilt for more complex, multi-step work

Build more reliable agents

We’re making it easier to create agents that work the way you expect from the start. We've strengthened how agents follow instructions and handle complex tasks, leading to more consistent results for your business scenarios. The new building experience streamlines Copilot Studio to bring the essential configurations to the forefront, reducing configuration tabs from nine to four. Together, these improvements help make it faster to build agents that are more reliable, more capable, and more efficient.

The new experience includes:

• New agentic orchestrator: Built on a new coding harness and CLI layer, the new orchestrator has stronger instruction adherence and long-horizon task execution. It supports recursive task execution, which enables the agents you create to be far better at working through complex and dynamic problems. It can also process large volumes of content and produce rich file outputs, opening up new document and data scenarios.
• New agent building interface: The new interface helps make building agents faster and more streamlined, allowing you to see your instructions, skills, tools and knowledge in one place. The new modern full page testing experience features better formatting and inline chain-of-thought and tool calling.
• Support for skills: Write your agent reusable instructions in markdown that can load on demand for completing specific tasks. You can also import skills you already have, including existing GitHub Copilot or Claude Code skills. 

Try the new agent building designer  in Copilot Studio.

Build more intelligent end-to-end workflows

Workflows are at the heart of reliable automation. With the new workflow designer, you can build intelligent, AI-driven processes in one place, combining structured workflow steps with agents. 

• New workflow designer: An intuitive visual designer for building agentic automations in a single, unified workspace. With node-by-node testing and robust versioning, it is now easier and faster to build, test, and publish running workflows. 
• Agent nodes: Call existing agents directly into a workflow. This allows you to leverage deterministic execution for reliability, while seamlessly handing off complex tasks that require greater intelligence and flexibility to agents within the workflow. 
• MCP servers: Connect to a broad ecosystem of model context protocol (MCP) server-enabled tools (preview). These allow workflows to execute tasks and involve users for approval, while staying within Microsoft security, permission, and compliance boundaries.