
Tech • AI • Robotics • Game
Claude cloud sessions let users start coding tasks on one device, continue them from another, and keep work running in a hosted virtual machine without leaving a local computer on.
A cloud session runs in a hosted virtual machine where Claude clones a repository from GitHub and works on its own branch. The session continues independently even if the user’s laptop is closed or asleep, allowing progress to continue across devices including desktop, browser, terminal, phone, and Slack.
Cloud sessions are positioned for longer-running tasks such as large refactors, migrations, or bug fixes that may outlast a normal work session. They are also meant for moments when a user wants to launch or continue development work away from a primary computer, including from a phone.
Because each cloud session uses its own branch, multiple tasks can be started at the same time without the user managing separate local worktrees. That makes it easier to hand off several jobs in parallel and review each result later.
To use cloud sessions, users must first connect their GitHub account and then install the Claude GitHub App with repository permissions. Private repositories appear only after the app has access to them, after which they become available when starting new sessions.
Sessions run by default in a cloud environment with common language runtimes, build tools, and databases preinstalled. Users who need more can create custom environments with configurable network access, environment variables, and setup scripts to install additional dependencies.
On Team and Enterprise plans, organizations can run cloud sessions on their own infrastructure. That gives larger customers a way to keep the workflow while aligning execution with internal infrastructure and policy requirements.
Each session starts from a fresh clone of the repository, so committed and pushed code is available immediately. But anything that exists only on a user’s local machine, such as personal settings, custom tools, plugins, or local configuration, does not carry over, making local execution better suited for tasks that depend on that setup.
Users can review code changes from a phone, leave comments that feed back into the next interaction, and create a pull request directly once the diff looks correct. After a pull request is opened, Claude can continue monitoring it and respond to failing checks or review comments.
Cloud sessions are distinct from Remote Control, which links the app to a Claude Code session running on the user’s own machine. In the cloud model, the work does not depend on the local computer staying powered on.
Users who prefer command-line workflows can run claude with the teleport flag to bring a cloud session onto their own computer and continue working there. That creates a bridge between hosted execution and local development when needed.
Cloud sessions extend Claude Code into an always-on, device-independent workflow built around hosted execution, GitHub integration, and branch isolation. The feature is aimed at making longer software tasks easier to launch, monitor, and review without tying progress to a single computer.
Ask a question