GitHub Copilot Remote Control: Monitor and Steer CLI Sessions from Your Phone

TL;DR

• GitHub Copilot CLI sessions can now be monitored and controlled remotely via github.com and GitHub Mobile
• Use /remote on to send any CLI or VS Code session to your phone — track progress, send instructions, approve changes
• Remote control works with any repository or directory, maintains full privacy, and enables complete workflows from mobile
• Available now in CLI (GA), VS Code, and JetBrains IDEs — no new install required if you're on the latest version

The Big Picture

Multi-agent workflows are becoming the norm. You kick off a refactor in VS Code, run tests in the CLI, scaffold a feature in the background. The problem? The moment you step away from your desk, you lose visibility into everything running. You can't check if that agent is still on track. You can't redirect it when it veers off course. You're stuck waiting until you're back at your laptop.

GitHub's new remote control feature solves this by making Copilot CLI sessions accessible from anywhere. Start a session locally, type /remote on, and it appears on github.com or the GitHub Mobile app. You get real-time monitoring, mid-flight corrections, and full workflow completion — all from your phone. This isn't just convenience. It's a fundamental shift in how agentic workflows can operate across surfaces.

How It Works

Remote control turns GitHub Copilot into a multi-surface platform. You start a session in VS Code or the CLI as usual. When you want to take it mobile, you run /remote on. That session immediately becomes accessible on github.com and the GitHub Mobile app. The session persists across all surfaces — CLI, VS Code, web, and mobile — with full state synchronization.

The feature works with any repository and even directories without repositories. You're not locked into a specific project structure. Once a session is remote, you get three core capabilities: real-time monitoring, mid-flight steering, and complete workflow execution.

Real-time monitoring means you see exactly what Copilot is doing as it happens. The plans it's researching, the files it's reading, the changes it's making, the commands it's running. Everything surfaces in the mobile app or web interface with no lag. You're not checking logs after the fact — you're watching the agent work.

Mid-flight steering lets you send additional instructions using natural language while a session is running. If an agent is heading in the wrong direction, you send a follow-up to redirect it. If you want to expand scope mid-task, you tell it. Approve or deny permission requests on the go. The agent responds to your instructions in real time, adjusting its approach without restarting the session.

Complete workflow execution means you can finish the entire development cycle from your phone. Plan and scaffold with Copilot CLI. Monitor progress in the GitHub Mobile app. Steer the session with follow-ups. Review the implementation plan and proposed changes. Create and review a pull request. Merge. All without returning to your laptop. The /remote on command removes the friction of switching surfaces.

Privacy is built in. Your sessions are only visible to you. No one else can see or access them. Remote control maintains full privacy across all surfaces.

What This Changes For Developers

Remote control fundamentally changes how you can work with agentic workflows. Before this, you were tethered to your desk. If you started a long-running agent task, you had to stay at your laptop or lose visibility. Now you can kick off a session, leave your desk, and stay in control from anywhere.

The practical impact shows up in several scenarios. You're commuting and want to check if that refactor finished. You're in a meeting and need to redirect an agent that's stuck. You're away from your desk and want to review and merge a PR that just got created. All of these are now possible without interrupting your workflow or waiting until you're back at your laptop.

The multi-surface continuity matters more than it sounds. You're not switching contexts or losing state when you move from CLI to mobile. The session persists with full history. You can start a task in VS Code, monitor it on your phone during lunch, send a follow-up instruction from the web interface, and return to VS Code later to review the final changes. The workflow stays intact across all surfaces.

This also changes how you think about agent supervision. You don't need to babysit agents at your desk anymore. You can let them run, check in periodically from your phone, and intervene only when needed. That's a different mental model than the current "start task, wait, check results" loop. It's closer to delegating work to a junior developer and checking in asynchronously.

For teams using Copilot Memory with user preferences, remote control adds another layer of flexibility. Your personalized agent behavior travels with you across surfaces, maintaining consistency whether you're steering from CLI or mobile.

Try It Yourself

If you're already using the latest version of GitHub Copilot CLI or GitHub Copilot in VS Code, there's nothing new to install. Start a session as you normally would, then use /remote on to send it to the web or mobile.

For new users, install GitHub Copilot CLI following the official installation guide. Once installed, start a session and enable remote control.

Download or update GitHub Mobile from the Apple App Store or Google Play Store to access remote sessions on your phone.

For detailed instructions, view the remote control documentation for CLI, VS Code, and JetBrains.

The Bottom Line

Use this if you run long-running agent tasks and need to stay mobile. The ability to monitor and steer sessions from your phone removes the desk-tethering problem that's plagued agentic workflows. Skip it if you're primarily doing short, synchronous coding sessions where you're always at your laptop — the overhead of enabling remote control won't add much value. The real opportunity here is unlocking asynchronous agent supervision. You can delegate more complex tasks to agents and check in periodically rather than babysitting them in real time. The risk is minimal — remote control is opt-in via /remote on, so it doesn't change your existing workflow unless you choose to use it. For developers already using GPT-5.3-Codex in Copilot Business or Enterprise, this adds another layer of flexibility to how you work with more capable models across surfaces.

Source: GitHub Blog