Claude Code cloud sessions have officially launched from research preview to general availability, giving developers a way to keep AI-powered coding work running on Anthropic's servers even after closing their laptop. On September 24, 2026, Anthropic announced the feature is ready for all users, along with a limited-time promotional credit: $100 for Pro subscribers and $250 for Max subscribers.

The announcement generated significant interest, with the initial post reaching approximately 654,000 views. However, many developers had immediate questions about what cloud sessions actually do, how they differ from existing tools, and whether the credit is worth claiming before the October 7 deadline.

What Exactly Is a Cloud Session?

A cloud session is Claude Code running on Anthropic's infrastructure instead of your personal computer. You provide the AI agent with a task and point it to your GitHub repository, and it continues working even after you close your laptop, lose internet connection, or step away from your desk. The practical advantage is straightforward: your machine is no longer the bottleneck.

This differs meaningfully from local sessions, where the code runs on your own computer and requires it to stay powered on. Cloud sessions also offer isolation, meaning the agent works in a hosted environment rather than directly accessing your machine's files, which limits potential damage if something goes wrong. Additionally, you can start a session on your desktop and check on its progress from your phone or web browser.

How Does the Credit Work, and When Does It Expire?

The promotional credit is separate from your regular usage limits, which is important if you've been rationing Claude Code near a weekly cap. Anthropic's framing is direct: "If you hit a limit locally, keep going in the cloud until your credit runs out." The credit applies automatically when you start a cloud session, but only if you've connected your GitHub account.

To claim the credit, you can either visit Anthropic's dedicated claim link or run the /claim-credit command in the Claude Code command-line interface (CLI). The deadline to claim is 11:59 PM Pacific Time on October 7, 2026, and the credit itself expires on November 4, 2026.

Steps to Get Started With Cloud Sessions

• Verify Your Plan: The credit is only available to existing Pro and Max subscribers; free tier users are not eligible.
• Connect GitHub: Cloud sessions require a GitHub connection to function, and the credit only applies once you've linked your account.
• Claim Your Credit: Use Anthropic's claim link or the /claim-credit CLI command before the October 7 deadline at 11:59 PM PT.
• Launch a Session: Start from claude.ai/code, the Code tab in the mobile app, the desktop app, or by running claude --cloud in your terminal.
• Monitor Expiration Dates: Keep track of both the claim deadline (October 7) and credit expiration (November 4) to avoid missing out.

When Should You Actually Use Cloud Sessions?

Cloud sessions are most valuable for specific workflows rather than everyday coding. They excel at long-running refactors and migrations where the work is mechanical and you want it completed overnight. They're also useful for parallel exploration, where you want several sessions trying different approaches to the same problem simultaneously. Travel and commute workflows benefit too, since you can kick off a task from your phone and let it run while you're away.

Cloud sessions also help if you've hit your local usage limits mid-task. The separate credit pool gives you a second source of capacity. They work well for reviews and fixes in continuous integration loops, and for any task that needs clear structure and planning.

However, cloud sessions are not the right tool if your codebase isn't hosted on GitHub, if your work requires access to local databases or hardware, or if your company policy prohibits sending code to hosted infrastructure. In those cases, Anthropic's Remote Control feature, which lets you control a local session from your phone or web browser, may be a better fit.

How to Write Prompts That Actually Work in Cloud Sessions

Cloud sessions reward a different approach to prompting than interactive chat. Because you're not watching the work happen in real time, clarity and structure become critical. Vague instructions that might work in a conversation with you watching will likely fail when the agent is working unsupervised.

• State the Finish Line: Be explicit about what "done" means. For example: "Done means all tests in packages/api pass, no new lint errors, and a PR description listing behavior changes."
• Bound the Scope: Clearly define which files or directories the agent can touch. Example: "Only touch files under src/billing/. If you need to change anything else, stop and write the reason to NOTES.md."
• Request a Running Log: Ask the agent to append what it tried and why to a SESSION_LOG.md file as it works. When you return, the log is the fastest way to review what happened.
• Require Evidence: Ask for test output or other proof for each claim that something is fixed, rather than accepting assertions.
• Set a Stop Rule: Define when the agent should give up. For example: "If two approaches fail, stop and summarize instead of continuing to try variations."

These rules transform a vague background job into something you can audit in five minutes when you return.

What Happens After the Credit Runs Out?

Anthropic has not yet announced pricing for cloud sessions after the promotional credit expires. This is an important gap to understand before you commit to the workflow. During the trial window, track how much credit each task consumes so you can estimate whether the feature will be worth paying for after November 4.

The company has also not provided detailed information about how cloud session usage is counted against your regular plan limits, so it's worth checking the documentation or contacting support to understand the billing mechanics before you rely on this feature for critical work.

Cloud sessions represent Anthropic's answer to a broader industry trend: every major coding tool is moving agents off the laptop and into hosted infrastructure. The design question is where the agent's compute and memory live, and who controls the sandbox. Anthropic's approach anchors the work to GitHub and runs it on Anthropic-hosted servers, giving developers a way to offload long-running tasks while maintaining some control over scope and access.