Microsoft opens its Copilot harness to developers through Foundry
Microsoft is making the Microsoft Copilot harness, the orchestration layer behind its own Copilot products, available through Microsoft Foundry. According to the company, developers and enterprise customers get a fully managed way to build agents that reason, use tools and complete complex, long-running tasks.
- Microsoft is making the Microsoft Copilot harness available through Microsoft Foundry as a fully managed option.
- The harness was built by GitHub engineers and powers Copilot Code, Copilot Studio, the GitHub Copilot app, the CLI and VS Code.
- According to Microsoft, it supports 20+ frontier models plus bring your own key and routes work to suitable models.
- Microsoft says porting the runtime to Rust cut session overhead and time-to-first token.
- Developers can download the SDK or run the harness in the cloud, while Microsoft runs the agent loop and sandbox.
What the harness is and where it runs
Microsoft describes the harness as everything around the model. If the model supplies the intelligence, the harness is the orchestration layer that gives an agent the right instructions, context, execution environment and controls to accomplish meaningful work. According to Microsoft, a good harness also supports open standards such as MCP and agent skills, as well as agent plugins.
The company says the harness was built by the GitHub engineering team and has since been adopted company-wide. It powers Copilot Code, Copilot Studio, the GitHub Copilot app, the CLI, VS Code and more. Microsoft argues that building a harness that works across multiple models and surfaces is hard because it has to balance performance, quality and cost, and then has to be maintained.
Multi-model routing and performance claims
Microsoft says the harness is built to perform across model families, with 20+ frontier models plus bring your own key. Rather than having customers keep up with and evaluate each new model, the company says it will route work to the right model or models based on the task. The pitch rests on the idea that the space moves quickly: a frontier model today might fall behind tomorrow, and building your own harness or model router to keep pace can be painful and slow.
On engineering, Microsoft says it recently ported the harness runtime to Rust. According to the company, this has resulted in industry-leading performance and reductions in both session overhead and time-to-first token. It also says the harness is designed to be fast, token-efficient and predictable, and that it has been proven at the scale of millions of users and hundreds of real-world use cases. These are the company's own claims, and the post does not include benchmark figures.
How developers can use it
Microsoft says the SDK can be downloaded, or the harness can be run in the cloud, fully managed by Foundry. Under this setup, Microsoft runs the agent loop and sandbox, with governance and optimization built in from day one. Customers bring the instructions, skills, tools and model choice.
The company stresses that its platforms remain open and interoperable, so developers can still bring in any harness and framework they prefer. Because agents built in Foundry can run on the same harness Microsoft uses across its products, Microsoft says its continuing improvements to the harness will benefit those agents too.
What it means for customers
For teams building agents, the offer is to skip the work of building and maintaining their own harness or model router and to use the one Microsoft and GitHub ship in their own products. Microsoft frames this as a step toward a stronger system around the model that preserves choice, and says it will keep developing the harness. Pricing and availability details are not given in the announcement.
