Skip to main content

Copilot Studio Agents

Copilot Studio gives business teams agents they own and update themselves. Three harnesses cover the work: multistep processes on the Microsoft Copilot harness, predictable conversations on the standard harness and your knowledge in Microsoft 365 Copilot Chat. Agents reach Teams, the web and enterprise apps through connectors, REST APIs and MCP servers.

Prompt or Hosted Agent

Microsoft Foundry has two kinds of agent. A prompt agent is the fastest path, run by Foundry from instructions, a model and tools with no code or infrastructure to manage. A hosted agent gives full control: it runs your own code and framework as a container, with a managed endpoint, autoscaling and a dedicated Entra identity from Foundry.

Proof of Concept

The proof of concept runs on your own use case in the Foundry agents playground, with every call traced and its quality measured in evaluations. Each iteration is saved as a version with rollback. Agents publish to Microsoft Teams and Microsoft 365 Copilot with their own Entra identity, private networking, role-based access control and content filters.
Team Updates It
the business team that owns the Copilot Studio agent updates it directly; every change stays inside the department
Full Code Control
your own code and framework run as a container; Foundry manages the endpoint, autoscaling and Entra identity
Fastest Path
the prompt agent runs from instructions, a model and tools; Foundry handles the code and infrastructure

From Copilot Studio to Microsoft Foundry

Copilot Studio Agents 2 steps
01

Three Harnesses for Three Scenarios

The Microsoft Copilot harness thinks through multistep business processes and adapts when circumstances change. The standard harness follows defined topics for predictable answers. The Copilot chat harness extends Microsoft 365 Copilot Chat with your own knowledge.
02

Distribute Across Channels and Tools

Agents can be distributed across Microsoft Teams, the web and enterprise applications. They connect to your systems through connectors, REST APIs and MCP servers, call workflows and hand tasks off to other agents.
Prompt or Hosted Agent 2 steps
03

Prompt Agent: The Fastest Path

You configure instructions, a model and tools, and Foundry runs the agent with no code or infrastructure. Portal-first, the agent takes shape interactively in the Foundry portal; code-first, it is defined through the SDK or REST API in your deployment pipeline.
04

Hosted Agent: Full Control

You bring your own code, built with the Microsoft Agent Framework, LangGraph, the OpenAI Agents SDK or a framework of your choice. Foundry provides a managed endpoint, automatic scaling and a dedicated Entra identity per agent.
Proof of Concept 2 steps
05

Test, Trace, Evaluate

In the agents playground, you chat with the agent or run it locally, with MCP servers testable directly. Every model call and tool call can be traced, and evaluations measure quality and catch regressions.
06

Version and Publish

Every iteration is automatically captured as a version, with rollback to any previous version and a comparison between versions. Published agents run through Microsoft Teams, Microsoft 365 Copilot and the Entra Agent Registry.

Frequently Asked Questions

When does a workload need the Microsoft Agent Framework?
When several agents coordinate across a long-running process, with state kept between steps and a person approving mid-run. A hosted agent on Microsoft Foundry fits a team of agents with its own release cadence inside one bounded service. The Microsoft Agent Framework combines AutoGen and Semantic Kernel, so code built on either has one framework to move to.
What is Microsoft Foundry retiring, and why does it matter now?
Microsoft Foundry retires its own workflows feature on 2026-12-01 and directs new work to the Microsoft Agent Framework. Every agentic workload on Foundry's workflow engine moves, planned now or forced by the deadline. Microsoft names the Agent Framework as the path for most of them. The Agent Framework is generally available and MIT licensed, a stable target for that move.
What does a proof of concept deliver?
A working agent on your own use case, tested in the Foundry agents playground. Every model call and tool call is traced. Evaluations measure its quality before it is published. The platform decision comes with it, written down with its reason.
What does the written decision look like in practice?
A short record per workload: the platform chosen, the reason at the time of the decision, the alternatives ruled out and the conditions that would change the answer. When a vendor retires a capability or a data residency rule changes, the record shows a reviewer exactly what to check.