A normal Claude Code session runs on your machine and stops when your machine does. A cloud session runs on cloud infrastructure instead, by default an isolated virtual machine that Anthropic manages. It keeps working after you close your laptop, and you can check on it or steer it from any device. The browser version at claude.ai/code is what the docs also call Claude Code on the web.
That buys you two things. You don’t need the repository checked out locally, and you can run several tasks at once, each in its own session on its own branch, without juggling worktrees.
How a session runs
You connect GitHub, pick a repository and branch, and describe the task. Claude Code clones the repository into a fresh VM, runs your setup script if you’ve configured one, and gets to work. At a stopping point it pushes a branch. You review the diff, leave comments on specific lines, send another message, or select Create PR. The session stays live after that, so follow-ups happen in the same conversation.
What the VM can reach is set by its cloud environment: network access level, environment variables and setup script. In Anthropic-hosted environments your GitHub credentials never enter the VM; a proxy attaches them on the server side.
Starting from anywhere
You can start a session in the browser, in the Code tab of the Claude mobile app, in the Desktop app, or from your terminal:
claude --cloud "Fix the authentication bug in src/auth/login.ts"That clones your current branch from GitHub, not your local checkout, so push first. Going the other way, claude --teleport (or /teleport inside a running session) pulls a cloud session into your terminal: it checks out the session’s branch and loads the conversation history. From the CLI the handoff is one-way; you can’t push an existing terminal session to the cloud.
Once a pull request exists, auto-fix can have Claude watch it and respond to CI failures and review comments, pushing a fix when one is clear and asking you when it isn’t.
The catch
Cloud sessions need a Pro, Max or Team plan, or an eligible Enterprise seat, and they share your account’s rate limits, so parallel tasks use them up faster. Cloning and pull requests require GitHub. Because auto-fix can reply to review comments under your GitHub account, check for comment-triggered automation before turning it on. If you want parallel work without leaving your machine, subagents split a task inside one local session. The cloud docs have the full reference.