Key takeaways
- Microsoft announced the Copilot harness on October 28, 2026, available through Microsoft Foundry.
- It supports 20+ models, bring-your-own-key, MCP, agent skills and agent plugins.
- The post gives no pricing, plans or general availability date.
Microsoft announced the Copilot harness on October 28, a runtime that lets developers run agents on more than 20 models through Microsoft Foundry. The announcement, by Foundry's Tina Schuchman and GitHub CTO Vladimir Fedorov, says GitHub's team built it and that Copilot products already run on it.
A harness is everything around a model that turns it into an agent. In the post's words, "the harness is everything around the model." The model supplies the intelligence. The harness supplies instructions, context, an execution environment and controls.
What developers get
Microsoft offers the harness two ways: as a downloadable SDK, in the github/copilot-sdk repository, or as a cloud-hosted harness managed by Foundry. In the hosted version, Microsoft runs the agent loop and the sandbox. Customers bring their instructions, skills, tools and choice of model.
It supports open standards named in the post: MCP, agent skills and agent plugins. It can route tasks to different models and accepts bring-your-own-key. Microsoft lists Copilot Code, Copilot Studio, the GitHub Copilot app, the Copilot CLI and VS Code as products that use it.
What Microsoft claims
Microsoft says it ported the runtime to Rust, which it says gives "industry-leading performance" and less session overhead. It also says the harness runs at the scale of millions of users. The post links to benchmark pages but gives no figures of its own, and we have not seen independent tests.
What we don't know yet
The post gives no pricing, no plan tiers and no model list beyond "20+". It also names no general availability date. For GitHub's own side of the multi-model push, see our report on Copilot Fusion. Watch the Foundry documentation for pricing.
