VS Code Agent Host Runs in the Background: Choose Phone Control, Local Machines, or Cloud Sessions
Local sessions can outlive editor windows, remote machines must stay running, and computer use remains experimental
Key takeaways
GitHub announced its VS Code Copilot Universe release on October 28, 2026. The new Agent Host lets sessions continue after a project folder or editor window closes, can run on another machine through Remote SSH or dev tunnels, and can be directed from GitHub Mobile. A local or remote machine and its Agent Host must remain running when the work executes there. Computer use is experimental and requires approval for each application. The open Agent Host Protocol lets custom clients reconnect to the same live session. This article compares local, remote, and cloud placement, then provides a first-run checklist for availability, permissions, desktop access, and recovery.
By ENHE AI · 2026-11-02 · AI-assisted drafting and editorial checks.
Direct answer: Execution location determines tools, uptime, and cost boundaries
GitHub announced the VS Code Copilot Universe release on October 28, 2026, and Agent Host lets work continue after a project folder or window closes. Local execution still depends on the machine, while cloud execution can continue after shutdown but enters cloud permission and compute boundaries.
How to choose local, remote, or cloud execution
Local Agent Host fits tasks that need local files, tools, or desktop apps. Remote SSH or a dev tunnel fits dedicated dependencies. The remote machine and its Agent Host must remain running. Cloud sessions run on GitHub infrastructure and can continue after a laptop sleeps, but repository permission and shared entitlement still matter.
Phone control and reconnection entry points
Run code agent host --github-environment from the project directory to monitor progress, steer work, or start a task from GitHub Mobile. The open Agent Host Protocol lets connected clients stay synchronized and catch up after reconnecting. The expected result is the same live session, not an untraceable copy.
Computer use and permission boundaries
Computer use requires approval for each application and remains experimental. Start with one application that has no sensitive data and record what it reads, clicks, and types. Organizations can disable it by policy. If control behaves unexpectedly, stop the session and revoke permission instead of retrying around approval.
Frequently asked questions
Will a local task always continue after VS Code closes?
Agent Host separates the session from a window, but the local or remote machine and Agent Host must still be running.
Does phone control automatically move work to the cloud?
No. The phone is a control surface; execution remains in the registered local, remote, or cloud environment.
Why this matters
The update separates an editor window from the agent runtime. That improves continuity but makes machine choice, network access, desktop permissions, and stopping behavior more important because background work can otherwise widen unseen impact.
Related tools and reading
These pages support the same task, workflow, or risk boundary.
- Continue a Copilot app session in VS Code
- Copilot local sandbox generally available
- Copilot cloud sessions and shared entitlement
Conclusion
Choose local Agent Host when local tools are needed and the machine can stay online, a remote machine for dedicated dependencies, and cloud execution when work must continue after shutdown. Start with one repository and one low-risk tool, then validate disconnect, reconnect, stop, and logs before expanding.
What this means for everyday users
Computer use requires approval for each application and remains experimental. Start with one application that has no sensitive data and record what it reads, clicks, and types. Organizations can disable it by policy. If control behaves unexpectedly, stop the session and revoke permission instead of retrying around approval.
Related tutorials
[[zh]][[/zh]][[en]]Review a public website[[/en]]
SearchScope SEO/GEO
[[zh]][[/zh]][[en]]Access and usage guide[[/en]]
AI Prompt Management System | 418 Bilingual Prompts for Writing, SEO and AI Creation
High-Frequency AI Prompts for Work, Learning, and Teaching guide
High-Frequency AI Prompts for Work, Learning, and Teaching
Related reading
GitHub Copilot Local Sandboxing Is Generally Available
GitHub made local sandboxing generally available on October 7, 2026 for GitHub Copilot CLI, the GitHub Copilot app, and VS Code sessions that use Agent Host. The feature applies operating-system controls to agent-run commands so users can restrict filesystem paths, network access, Git credentials, and supported local tools. GitHub's documentation also sets important boundaries: local sandboxing is off by default, CLI and app settings are separate, remote MCP servers are not sandboxed, and some in-process file tools enforce policy on a best-effort basis rather than through the operating-system sandbox. This guide focuses on verification rather than the toggle itself: enable the feature, test denied reads and writes, test blocked network access, confirm credential masking, inspect MCP placement, and configure fail-closed behavior where the organization requires enforcement.
GitHub Copilot Sessions Can Move to VS Code With Context, Diffs, and Worktree Intact
GitHub has documented a handoff from a Copilot app session to Visual Studio Code. The workflow keeps the session context, changed-file diff, and the same isolated worktree, so a developer does not need to rebuild the task history in a new chat. This article explains the prerequisites and provides a practical review sequence: open the session, inspect the diff, edit locally, run the relevant tests, and decide separately whether to push. The feature is useful when agent work needs human judgment or local tooling, but continuity is not proof that the generated changes are correct. Repository permission, the same GitHub account in both applications, and an existing session with changes are still required.
GitHub Copilot Cloud Sessions Are GA: How Cross-Device Work Uses Actions Entitlement
GitHub made its connected Copilot cloud experience and cloud sandboxes generally available on paid Copilot plans on October 28, 2026. A user can start a task on github.com, GitHub Mobile, Copilot CLI, the Copilot app, VS Code, Slack, or Microsoft Teams, then continue the same session elsewhere. The included Actions compute entitlement becomes one shared, dollar-based pool for standard hosted runners, cloud sandboxes, and certain larger runners. Billing for cloud-sandbox usage begins in November 2026, so cross-device convenience can affect the same budget used by builds. This article helps users check repository scope, write access, tool permissions, entitlement consumption, and stop conditions before their first cloud session.
From Chat Boxes to Personal AI Companions: AI Assistants Are Entering the Desktop Execution Era
AI assistants are moving from answering questions toward continuing real tasks. AI agents, MCP tool ecosystems, personal memory, and local workbenches are pushing this shift together. For users, the real value is not another chat box, but less repeated context setup and more continuity from thinking to doing.
What Is Replica? A Validation Guide for AI Prototypes Built From Your Design System
Replica is an open-source AI prototyping tool released by GitHub Design on October 28, 2026. It uses a team’s existing components, patterns, and documentation to guide interactive prototypes instead of starting from a generic visual language. Team members can describe and refine a prototype through the GitHub Copilot app, begin locally, and optionally deploy a self-hosted Replica Hub for review and comments. The project is available under the MIT license. This durable guide explains the required design-system inputs, a small example task, expected outputs, consistency checks, feedback handling, and failure recovery so a prototype can inform product discussion without being presented as production code or a tested interface.
GitHub Issue Artifacts Preview: What Agent Outputs Can Store and How Version History Works
GitHub announced Issue Artifacts in public preview on October 28, 2026. Teams can attach agent-generated plans, research, links, visualizations, and interactive HTML directly to an issue. Updates made on github.com retain version history as an artifact changes. GitHub CLI adds gh issue artifact commands to create, edit, list, view, download, and delete artifacts, while Markdown artifacts can include image and video attachments. This explainer uses a root-cause-analysis example to show which outputs belong beside an issue, how to separate observed evidence from unverified conclusions, when history helps reviewers follow a decision, and when an incorrect or sensitive artifact should be corrected or removed rather than being treated as repository truth.