Microsoft is making the Copilot harness available through Microsoft Foundry. The company says developers and enterprise customers can use it to build agents that reason, use tools and carry out long-running tasks across different AI models.
A harness is the software around an AI model that manages an agent’s instructions, context, execution environment and controls. Microsoft says its harness supports open standards including MCP and agent skills, as well as agent plugins.
Built by GitHub engineers, the harness already powers Copilot Code, Copilot Studio, the GitHub Copilot app, the CLI and VS Code, according to Microsoft. The company says it has adopted the harness across its Copilot products.
The harness works with more than 20 frontier models
Microsoft says the harness works across model families, with 20-plus frontier models and a bring-your-own-key option. It can route work to a model or models based on the task. Developers can also choose a model themselves.
Microsoft runs the agent loop and sandbox, with governance and optimization built in, while developers supply the instructions, skills and tools. The company says it recently moved the harness runtime to Rust, reducing session overhead and the time before a model starts responding.
The SDK is available to download or run in the cloud
The SDK is available to download, and developers can also run the harness in the cloud, managed by Foundry.
