Author(s): Dave R – Microsoft Azure & AI MVP☁️

Originally published on Towards AI.

A technical guide to running the Copilot CLI from your phone, in an isolated cloud environment, and across a history you can query.

GitHub Copilot has grown past the editor, and the command-line agent can now run in three places that used to be off-limits: steered from your phone or a browser while it keeps executing on your laptop, running entirely inside an isolated cloud environment that survives your machine shutting down, and drawing on a searchable history of everything you have done across every Copilot surface.

After establishing Copilot CLI as a terminal-native, agentic, steerable environment (including built-in MCP support and the fact it’s one surface among many), the article explains three capabilities in turn: remote control, which streams input/output so you can monitor and steer a running local interactive session from GitHub Mobile/web without “ordinary syncing”; cloud sandboxes, which launch an isolated, ephemeral Linux environment via a simple flag (with policy inheritance and snapshot/resume on idle so sessions survive laptop shutdowns); and Chronicle, which indexes your session history across Copilot surfaces into a queryable memory layer with commands for standups, tips, improvements, search, and cost governance. It then summarizes how these features fit together in a layered architecture (surface, execution, memory, identity/policy) and closes with setup guidance and a recommended way to adopt each capability safely.

Read the full blog for free on Medium.

Join thousands of data leaders on the AI newsletter. Join over 80,000 subscribers and keep up to date with the latest developments in AI. From research to projects and ideas. If you are building an AI startup, an AI-related product, or a service, we invite you to consider becoming a sponsor.

Published via Towards AI