Microsoft Gives Copilot Local File Context and OS-Wide Actions

Microsoft on October 7, 2026 announced that Copilot on Copilot+ PCs will gain local context from a user’s files and recent activity, the ability to take actions across Windows, and access to AI models running on the device, in a Windows Experience Blog post laying out what the company calls hybrid intelligence.

The announcement, from Pavan Davuluri, Executive Vice President of Windows + Devices, extends the new Copilot Microsoft introduced on September 25, 2026, which is built around three experiences: Home, Code, and Autopilot.

Microsoft described three local capabilities. With local context, and with the user’s permission, Copilot can understand relevant content on the PC, including files and recent activity. With local actions, Copilot can work on the user’s behalf across the operating system, including organizing files, assessing device diagnostics, troubleshooting issues, coding, and carrying out workflows directly on the device. With local models, Copilot can tap AI models running on the PC, using local compute for the right tasks while combining it with cloud intelligence when needed.

Microsoft said Home can use local context to pull in files the user has been working on and turn them into collaboration-ready artifacts, while Code gains the ability to build native Windows apps from a single prompt, run code more securely with Microsoft Execution Containers support, and use local models to manage token costs. Autopilot, which Microsoft describes as a persistent, proactive and personal agent, gains local context, local models, and PC actions in its agent flow. Copilot features powered by hybrid intelligence are expected to begin rolling out on Copilot+ PCs in the coming months, Microsoft said, with availability and timing varying by device, market, and silicon platform.

Local Models and HydraFusion Routing

Microsoft said it is bringing MAI Code 1.1 Flash, a coding model introduced at Build with 137 billion total and 6.8 billion active parameters, to devices using 3-bit precision, a change the company said reduces model size by nearly 80% while preserving coding quality and supporting a 256K context window locally. It also said an upcoming NVIDIA Nemotron model with over 70 billion parameters, quantized in 2-bit to use just over 20GB of memory, and DeepSeek V4 Flash, a 284-billion-parameter model, are being brought to run locally on RTX Spark hardware.

GitHub’s HydraFusion, which routes each task to the right model in the cloud, is being extended to Windows so it can also tap into models running locally on device. Hybrid intelligence on Windows powered by HydraFusion is coming to the GitHub Copilot app, GitHub Copilot CLI, and Visual Studio Code in experimental preview later in October 2026, Microsoft said. The company also announced llama.cpp support in Windows ML, its runtime for deploying models across GPU, NPU, and CPU, giving developers more open-source model choices.

Microsoft framed the local push around cost, stating that customers’ needs are outpacing what their cloud budgets can support and that hybrid intelligence is meant to stretch those budgets while preserving access to frontier capabilities. The company said more than 2 trillion inferences are run locally each month across all Copilot+ PCs, and that over 40% of laptops being built for business are Copilot+ PCs.

Microsoft Execution Containers Reach General Availability

Microsoft Execution Containers (MXC) became generally available on Windows 11 on October 7, 2026, enabling organizations to define which files and networks agents can access, with those policies enforced at runtime. According to a companion Windows Developer Blog post by Logan Iyer, Corporate Vice President of Windows Platform + Developer, the containment layer keeps policy outside the agent workload’s control, so the agent or its generated code cannot grant itself additional access.

Developers declare the resources a workload requires, such as files and network destinations, through a unified JSON configuration schema and multi-language SDK, and MXC maps the requested controls to containment backends on Windows, macOS, and Linux. Containment options range from process isolation to a session container, WSLc, virtual machines, and Windows 365 for Agents. Only Windows supports the session container, which runs an agent in a separate OS-isolated session with its own local agent identity and isolated desktop, clipboard, UI, and input boundaries.

MXC supports Enforcement, Learning, and Permissive modes, and only on Windows can MXC process containers produce an agent activity report showing which resources a workload attempted to use, to help craft least-privilege policies. Microsoft’s stated security model pairs that containment with identity, so IT can tell which agent acted, and manageability, connecting agent activity into tools such as Agent 365 and Intune.

Agents already supporting MXC include OpenAI’s Codex, GitHub Copilot, OpenClaw, Replit, LM Studio, NVIDIA’s OpenShell, and Unsloth AI, with Anthropic Claude Code, Box, Egnyte, Heidi Health, Hermes Agent by Nous Research, Manus, Perplexity, Raycast, and Simular among those planning support, Microsoft said. Meta’s Muse personal AI agent is coming soon as a native Windows app with MXC integration. Windows will also soon enable Microsoft Entra to distinguish agent activity from user activity in Microsoft Agent 365, and Intune policy will soon be available to manage MXC process containers on Windows 11.

Actions in Windows Search and New Devices

Microsoft introduced a new Windows Search experience with thousands of actions available directly from the taskbar, beginning with tasks such as switching to dark mode, turning on Do Not Disturb, enabling Airplane mode, dimming the screen, muting audio, minimizing or arranging windows, and sending a text message. It will also begin adding an opt-in option to integrate the new Copilot with Search on the taskbar, providing quick answers inline with the ability to jump to the full Copilot app, first rolling out to select markets later in 2026. The new actions began rolling out to Windows Insiders in the experimental channel on October 7, 2026.

The announcements came at Microsoft’s October 2026 Windows and Surface event, where the company also opened pre-orders for Windows PCs powered by NVIDIA RTX Spark from ASUS, Dell, HP, Lenovo, MSI, and Microsoft Surface. Microsoft said the Surface Laptop Ultra, built around RTX Spark with up to 128GB of unified memory, can run AI models exceeding 120 billion parameters locally; it becomes available October 16, 2026, and the Surface RTX Spark Dev Box ships to U.S. customers in November. Deskside systems powered by NVIDIA DGX Station, which Microsoft said can run frontier-class models at more than a trillion parameters locally, are due on Windows later in 2026, including the Dell Pro Precision with GB300 and the HP ZGX Fury AI Station.