Microsoft is releasing its internal Copilot harness to developers and enterprise customers via Microsoft Foundry, providing a managed platform to build multi-model AI agents that handle complex, long-running tasks across tools and surfaces, blogs.microsoft.com reported on October 28, 2026.
The Engineering Burden of Multi-Model Orchestration
Building an AI agent that works reliably across multiple frontier models is notoriously difficult. Developers must constantly solve for performance degradation, variable output quality, and spiking operational costs. Worse yet, maintaining that infrastructure as new models enter the market drains valuable engineering resources.
GitHub has released Copilot Fusion in GitHub Copilot, which orchestrates between Anthropic and OpenAI models. It scores 2 points below Opus 5.5 in Claude Code at half the Cost per Task in the Coding Agent IndexGitHub describes Copilot Fusion as a runtime orchestration system,… pic.twitter.com/gx3ZJ6Hm86
To stay competitive, organizations must tap into diverse model capabilities without constantly rewriting their underlying orchestration layer. The GitHub engineering team faced this exact challenge. Their solution was to build a dedicated, multi-model harness that now powers corporate surfaces including Copilot Code, Copilot Studio, the GitHub Copilot app, the command-line interface, and Visual Studio Code.
Architecture and Performance Upgrades in Rust
The harness acts as the crucial execution environment surrounding the core intelligence model. It supplies agents with instructions, strict context windows, execution controls, and open standards support like the Model Context Protocol (MCP), agent skills, and plugins.
Microsoft recently ported the entire harness runtime to Rust. This systems-level rewrite delivered significant performance gains, notably slashing session overhead and reducing time-to-first-token latency. The platform now routes workloads intelligently across more than 20 frontier models—while allowing developers to bring their own API keys—ensuring each task runs on the optimal model family without requiring manual stack rewrites.
Enterprise Integration and Managed Deployment via Foundry
By making the harness available through Microsoft Foundry, the company is opening up the exact production environment used by millions of developers. Enterprises can deploy agents that execute multi-step workflows while Microsoft manages the underlying agent loop, secure sandboxes, governance protocols, and optimization routines.
Developers retain total freedom over instructions, custom tools, and model selection. Because Microsoft continuously refines the core harness across its internal product suite, improvements roll out automatically to custom agents built within Foundry.
Open standards remain foundational to the release, ensuring teams can integrate external frameworks and alternative harnesses alongside the new Microsoft runtime.
