From the source
Microsoft has made the Copilot harness, the multi-model orchestration layer behind its Copilot products, available to developers and enterprises through Microsoft Foundry.
- A companion Foundry Blog post introduced it in Foundry Agent Service in public preview alongside general availability of long-running resilience for hosted agents.
- Microsoft describes a harness as everything around the model: the orchestration layer that supplies instructions, context, an execution environment and controls, with support for open standards such as the Model Context Protocol plus agent skills and plugins.
- The
GitHub engineering team built the harness and Microsoft has adopted it company-wide; the company says it performs across more than 20 frontier models with bring-your-own-key and routes work to the right model based on the task.
- During the preview, customers select the harness when creating an agent, existing agents are unchanged and switching the harness after creation is not yet supported.
- Billing starts December 1, 2026 on a consumption basis: runtime usage is free before then, and from that date sandbox compute costs $0.0994 per vCPU-hour and memory $0.0118 per GiB-hour, with prices varying by region and agreement.
- Long-running resilience targets work that takes hours or days: Foundry detects interruptions and restarts with the original input, while customer code saves progress and must prevent duplicate actions when updating other systems.