Microsoft opens Copilot harness to Foundry for multi-model agents

News

Microsoft said its Copilot harness is now available through Foundry, bringing a managed runtime for multi-model agents and tool use.

Microsoft said on Oct. 28, 2026, that it is making the Microsoft Copilot harness available through Microsoft Foundry, extending a runtime it says already powers Copilot Code, Copilot Studio, the GitHub Copilot app, the CLI and VS Code. The company framed the move as a way for developers and enterprise customers to build agents that can reason, use tools and complete longer-running tasks without switching harnesses or the rest of their stack.

Evidence and context

In Microsoft’s account, the harness is the orchestration layer around an AI model: it supplies instructions, context, execution environment and controls. The company said a harness suitable for this work also needs support for open standards such as MCP, agent skills and agent plugins. Microsoft described the field as fast moving and said maintaining a custom router or harness can be difficult because performance, quality and cost must be managed continuously.

Microsoft said the Copilot harness was built by GitHub engineers and has been standardized across its Copilot experiences. The company said the system is designed to work across model families, including “20+ frontier models” plus bring-your-own-key support, and that it automatically routes work to the right model or models for a given task. Microsoft also said it recently ported the harness runtime to Rust, which it claims improved performance and reduced both session overhead and time-to-first token.

The company did not provide independently verifiable benchmarks in the post. It said the harness is battle-tested across “millions of users” and “hundreds of real-world use cases,” but those figures were not broken out further. Microsoft also said Foundry customers will be able to run the harness in the cloud as a fully managed service or download the SDK, while continuing to use other harnesses and frameworks if they choose.

What the update means for Copilot and Foundry

The announcement follows earlier Microsoft and GitHub posts dated Aug. 3, Sept. 2 and Sept. 25, 2026, which described newer harness capabilities, connected agents and Copilot surface updates. In the background material supplied with this article, Microsoft distinguishes the GitHub Copilot harness from the GitHub Copilot service itself and says Copilot Studio agent data is not sent to that service. It also says existing Copilot Studio privacy, security, compliance and residency commitments continue to apply.

Microsoft’s guidance also cautions that the harness is best suited to longer, context-heavy work and may add unnecessary latency and Copilot Credit consumption for simpler tasks. The company recommends evaluating scenarios individually rather than automatically migrating every workflow. That makes the new Foundry availability a platform expansion, not a blanket replacement for the standard harness or other execution paths.

Microsoft said the overall goal is a “stronger system around the model that preserves choice.” The company’s post did not give a release schedule for new product features beyond the availability of the SDK and cloud-hosted option through Foundry.