Understanding Copilot Studio Harnesses and Their Costs

• An AI harness is a layer between an agent or an AI application and the models, tools, and services it uses.
• Copilot Studio has added support for multiple AI harnesses with distinct capabilities, behaviors, and licensing models.
• Harness selection affects costs of building and using Copilot Studio agents and workflows.
• Organizations can’t directly specify which harnesses are available, but they can use governance features to indirectly control harness usage.

Copilot Studio supports three AI harnesses: Copilot Chat, Standard, and GitHub Copilot. Design choices made by makers using Copilot Studio when building an agent or workflow determine the harness. Organizations using or evaluating Copilot Studio should understand the differences between the harnesses and make governance decisions about which can be used because the choice of harness affects costs.

What Is an AI Harness?

An AI harness is the framework (sometimes called a scaffold) that sits between an agent’s business logic and the underlying AI platform being used. It provides runtime services such as model orchestration, access to tools, authentication, reasoning workflows, knowledge retrieval, governance controls, and billing.