Microsoft Makes Windows 11 a Hub for Local and Cloud AI Agents

Microsoft is turning Windows 11 into a platform for AI agents that split work between local devices and cloud services, combining on-device processing with new controls for security and management.

Microsoft Execution Containers became generally available on Windows 11 on Oct. 7, 2026. The system lets organizations define which files and networks an agent can access, with those policies enforced while the agent runs.

“The policy remains outside the agent workload’s control, so the agent or generated code cannot grant itself additional access,” Microsoft said.

Execution Containers are also generally available across Windows, macOS and Linux, with support for Windows 365 also generally available. Microsoft separately described identity and management features for distinguishing agent activity and governing it through Agent 365 and Intune.

“That’s why we’re building Windows as the home for hybrid intelligence: a platform where agents can run locally when it makes sense, reach the cloud when they need to, and operate with the security and manageability organizations expect,” Pavan Davuluri, Microsoft’s executive vice president, Windows + Devices, said.

The company is pairing those controls with smaller local models intended to reduce hardware and memory demands. MAI Code 1.1 Flash has 137 billion total parameters and 6.8 billion active parameters. Using 3-bit precision reduces the model’s size by nearly 80%, while its local context window reaches 256K.

Microsoft also identified an upcoming NVIDIA Nemotron model with more than 70 billion parameters and DeepSeek V4 Flash with 284 billion parameters for local operation on RTX Spark hardware. The Surface Laptop Ultra and other Windows PCs using NVIDIA RTX Spark are available for pre-order, with the Surface model offering up to 128 GB of unified memory and the ability to run AI models exceeding 120 billion parameters locally.

Copilot on Copilot+ PCs is expected to gain permission-based access to local context, local actions and local models in the coming months. Microsoft has not provided a specific release date.

Hybrid intelligence is scheduled to enter experimental preview in the GitHub Copilot app, GitHub Copilot CLI and Visual Studio Code later in October 2026. The approach extends Microsoft’s earlier work on local coding-model inference in GitHub Copilot, where users are expected to switch between cloud-based and local models.

The strategy is intended to improve security, manageability, responsiveness and AI-token efficiency. No quantified estimate was provided for cloud-cost savings, lower customer spending or the effect on Windows revenue.