Claude Code goes cloud: Run long tasks even with your laptop shut

Running AI coding tasks from your local terminal has always had one annoying friction point: you had to keep your laptop awake and connected. If you closed the lid to head home or your Wi-Fi dropped, the entire half-finished refactor was gone.

Anthropic solved this on September 24 by launching cloud sessions for Claude Code. Basically, it moves your coding agent from your local hardware into managed cloud sandbox environments on Anthropic’s infrastructure.

You can now kick off a task with `claude --cloud`, and Claude will execute it remotely without relying on your machine. You can shut your laptop, walk away, and it will keep coding in the background. The feature is open to Pro, Max, Ultra, Team, and Enterprise subscribers. According to reports, Anthropic is also handing out 100 dollars in dedicated credits for Pro users and 250 dollars for Max users to get them started, with no extra credit purchases required for subscribers.

Offload from CLI, monitor from your phone, teleport back home

What makes this setup genuinely useful is how cleanly it bridges your local shell, the cloud, and mobile devices.

You can interact with sessions through claude.ai/code in your browser, the Code tab in the Claude mobile app, the desktop app, or straight from the command line. Running `claude --cloud` spawns an independent session in the cloud. You can fire off several tasks simultaneously, and each runs in its own cloud container without chewing up local CPU or RAM.

While away from your desk, you can monitor output or steer agents directly from your phone. When you sit back down at your machine, pulling the session into your local terminal is a matter of running `claude --teleport`. It works like an instant handoff, bringing the entire conversation context right back to your prompt.

If your local repository is not pushed to GitHub or lacks the Claude GitHub App, the CLI automatically bundles your local files up to 100 MB and pushes them to the cloud sandbox. Note a small platform quirk: on macOS and Linux, the tool automatically excludes uncommitted sensitive files like .env or private keys, but on native Windows it currently uploads everything as is, meaning you have to stash secrets manually. Also, teams with Zero Data Retention or HIPAA compliance turned on cannot use cloud sessions yet, nor can users connecting via Amazon Bedrock or Google Cloud.

Why it matters

Moving an agent from local terminal execution to fully managed background instances feels like the exact right evolution for coding assistants.

Most modern AI coding tools still behave like copilots in the strictest sense: they sit in your editor, hold onto your terminal, and pause the second you close your laptop. Claude Code cloud sessions shift the dynamic toward an asynchronous junior engineer. You assign a time-consuming test suite or boilerplate migration at the end of the day, close your lid, and inspect the finished PR or teleport the session back the next morning.

For developers, the parallel execution alone justifies the workflow. Running multiple cloud tasks concurrently without melting local fans turns long-running tasks into background background noise. With bonus credits included, there is virtually zero friction to try it out.

Across the industry, this moves the competitive bar beyond raw benchmark scores and autocomplete speed. The real battlefield is workflow integration and autonomy. There are still rough edges to polish, particularly around secret sanitation on Windows and enterprise compliance hurdles. But overall, the seamless round-trip between local terminal and persistent cloud environments sets a high standard for how autonomous developer tooling should look.