Microsoft and GitHub have made their Copilot harness generally available to developers and enterprise customers through Microsoft Foundry. The harness is an orchestration layer that enables building AI agents capable of reasoning, using tools, and completing long-running tasks, and it already powers products such as Copilot Code, Copilot Studio and the GitHub Copilot app.
The Copilot harness is the orchestration system that surrounds an AI model, providing instructions, context, execution environment and controls. According to Microsoft, building and maintaining such a system internally is difficult and slows down development, especially as frontier models evolve rapidly.
The harness was originally built by GitHub engineers and has been battle-tested at the scale of millions of users across Microsoft's own Copilot experiences. It supports more than 20 frontier models and allows customers to bring their own model keys. The company has also ported the harness runtime to Rust, achieving what it described as industry-leading performance with reductions in session overhead and time-to-first token.
The harness supports open standards including MCP and agent skills, and is flexible enough to work with any application, agent or service. By offering the harness as a managed service through Microsoft Foundry, the company said customers can focus on building instructions, skills and tools while Microsoft runs the agent loop and sandbox with built-in governance and optimization.
Developers can download an SDK or run the harness in the cloud through Foundry. Microsoft emphasised that its platforms remain open, allowing developers to continue using any other harness or framework if they prefer.
Analysis
Why This Matters
- Enterprise developers and organisations can now adopt a production-tested multi-model orchestration system without having to build and maintain their own.
- The move could accelerate the deployment of AI agents in business settings by lowering the technical barrier to entry.
- Because Microsoft and GitHub are standardising on the same harness, continuous improvements will flow to all products and customers that use it.
Background
AI agents need more than a large language model; they require an orchestration layer to manage instructions, context, tools and execution. Building a high-quality, multi-model harness that stays current is a significant engineering challenge. Microsoft and GitHub had already solved this for their own internal use and are now making the solution available as a managed service through Microsoft Foundry.
What to Watch
- Adoption of open standards such as MCP within the harness and the broader ecosystem.
- Performance benchmarks from early users, particularly around the Rust-based runtime.
- Competitive responses from other cloud platforms and orchestration providers.
Checked
12 claims in this story checked against 1 source: 11 supported.
Show each claim
- SupportedMicrosoft has opened the Copilot harness to enterprise developers through Microsoft Foundry.The Official Microsoft Blog: “And today, we’re making that harness available through Microsoft Foundry. That gives developers and enterprise customers a fully managed way to build agents”
- SupportedThe harness was built by GitHub engineers.The Official Microsoft Blog: “Luckily, the GitHub engineering team built a best-in-class, truly multi-model harness”
- SupportedThe Copilot harness is an orchestration layer for agents.The Official Microsoft Blog: “It’s the orchestration layer that gives an agent the right instructions, context, execution environment, and controls to accomplish meaningful work.”
- Partly supportedThe harness is generally available to developers and enterprise customers through Foundry.The Official Microsoft Blog: “And today, we’re making that harness available through Microsoft Foundry. That gives developers and enterprise customers a fully managed way to build agents”
- SupportedThe harness enables agents to reason, use tools and complete long-running tasks.The Official Microsoft Blog: “That gives developers and enterprise customers a fully managed way to build agents that reason, use tools, and complete complex, long-running tasks.”
- SupportedThe harness powers Copilot Code, Copilot Studio and the GitHub Copilot app.The Official Microsoft Blog: “The Microsoft Copilot harness powers Copilot Code, Copilot Studio, the GitHub Copilot app, the CLI, VS Code, and more.”
- SupportedMicrosoft says building and maintaining an internal harness is difficult and slows development as frontier models evolve.The Official Microsoft Blog: “A frontier model today might fall behind tomorrow. And building your own harness or model router to keep up can be painful and slow.”
- SupportedThe harness has been tested at the scale of millions of users across Microsoft’s Copilot experiences.The Official Microsoft Blog: “And because the Copilot harness powers surfaces across Microsoft, it’s battle-tested and proven performant at the scale of millions of users and hundreds of real-world use cases.”
- SupportedThe harness supports more than 20 frontier models.The Official Microsoft Blog: “The harness is built to perform across model families, with 20+ frontier models plus bring your own key, so you can use the right model for the right job.”
- SupportedCustomers can bring their own model keys.The Official Microsoft Blog: “The harness is built to perform across model families, with 20+ frontier models plus bring your own key, so you can use the right model for the right job.”
- SupportedMicrosoft ported the harness runtime to Rust, which resulted in industry-leading performance and reductions in session overhead and time-to-first token.The Official Microsoft Blog: “We recently ported the harness runtime to Rust, which has resulted in industry-leading performance and reductions in both session overhead and time-to-first token.”
- SupportedMicrosoft runs the agent loop and sandbox, with governance and optimization built in.The Official Microsoft Blog: “With the Copilot harness, Microsoft runs the agent loop and sandbox, with governance and optimization built in from day one.”
