Claude Code Cloud Sessions Are Generally Available: How to Claim the $100 / $250 Credit and When to Use Them

Claude Code cloud sessions are out of research preview. On September 24, 2026, the ClaudeDevs account announced that cloud sessions are "officially available" and "let you keep Claude Code working, even when your laptop is closed." To get people to try them, existing subscribers get a one-time credit: $100 on Pro, $250 on Max.

The announcement drew about 654K views, and the replies asked the questions most developers will have: is this new, how is it priced, and how is it different from what I already use? This guide answers those, walks through claiming the credit before the October 7 deadline, and explains when a cloud session is the right tool. It builds on our coverage of Claude Code Projects and running Claude Code from your phone.

TL;DR: Claude Code cloud sessions

What a cloud session actually is

A cloud session is Claude Code running on Anthropic's machines instead of yours. You hand it a task and a GitHub repository, and it keeps working after you close the lid, lose Wi-Fi or leave the room. Anthropic's pitch: "Cloud sessions run on Anthropic-hosted infrastructure, so the work keeps going even without your computer running."

The practical differences from a local session:

• Your laptop is not the bottleneck. No sleep, no battery, no VPN drop killing a long task.
• Isolation. The agent works in a hosted environment, not on your machine's files, which limits blast radius.
• Accessible everywhere. Start on desktop, check on mobile.
• It draws from credit, not just plan limits. The one-time credit applies automatically when you start a session, and it is "separate from your usage limits."

That last point matters if you have been rationing Claude Code near a weekly cap. Anthropic's own framing: "If you hit a limit locally, keep going in the cloud until your credit runs out." For context on how limits have changed across tools, see Claude, Codex and Cursor usage limits reset.

How to claim the credit, step by step

1. Confirm your plan. The credit is for existing Pro and Max subscribers: $100 on Pro, $250 on Max.
2. Connect GitHub. You need it to start a session, and the credit only applies once you do.
3. Claim. Open Anthropic's claim link, or run /claim-credit in the Claude Code CLI. The graphic shows the link path as claude.ai/code/claim-credit/10.
4. Start a session. From claude.ai/code, the Code tab in the mobile app, the desktop app, or claude --cloud.
5. Watch the dates. Claim by 11:59 PM PT on October 7. The credit expires 11:59 PM PT on November 4.

One credit per account, cloud sessions only, terms apply. If you already use cloud sessions from the research preview, the announcement's point is that the credit is a bonus to encourage more use.

Cloud sessions vs Remote Control vs Projects

Replies to the announcement showed real confusion, so here is the clean distinction.

One reply complained that cloud sessions are useless compared with a Codex-style remote control that runs on your own PC, because their project is huge and does not use GitHub. That is a legitimate boundary: if your codebase is not in GitHub or depends on a local environment, Remote Control is the closer fit. Our mobile and Remote Control guide covers that path, and running coding agents from your phone compares the alternatives.

Another reply asked how Projects differ. Projects, which rolled out on September 17, organize work as a coordinator conversation that starts a cloud-session thread for each piece. Cloud sessions are the substrate; Projects are the orchestrator on top.

What people are asking

"I thought this was free. How will pricing work?" The credit is finite and one-time. Anthropic has not said in this announcement what usage costs after the credit is spent, so plan for it: check how cloud session usage is counted against your plan, and track spend during the trial window.

"Is this new? I've been using cloud all along." Cloud sessions existed in research preview. The change is general availability plus the credit. If you used them before, nothing dramatic changes except stability expectations and the promotion.

"Don't Projects already run in the cloud?" Yes, Projects use cloud threads. The announcement is about the underlying feature reaching GA.

"My repo isn't on GitHub." Then cloud sessions are not a fit yet. GitHub connection is required.

When cloud sessions are worth it

Good fits:

• Long-running refactors and migrations where the work is mechanical and you want it done by morning.
• Parallel exploration: several sessions trying different approaches to the same problem.
• Travel and commute workflows: kick off from the mobile Code tab.
• Hitting local limits mid-task: the credit gives you a second pool.
• Reviews and fixes in CI-like loops, especially with Ultraplan for planning.

Poor fits:

• Tasks that need your local databases, hardware, or private network.
• Work you want to watch and steer keystroke by keystroke.
• Anything that violates your company's rules about sending code to hosted infrastructure. Check policy first.

How this fits the 2026 agent trend

Every major coding tool is moving agents off the laptop. Cursor added cloud agents on your own infrastructure, event-driven cloud agents and Projects. OpenAI's ChatGPT Work runs cloud browsers and tasks. The design question is where the agent's compute and memory live, and who controls the sandbox. Anthropic's answer is hosted infrastructure with GitHub as the anchor. For a survey of the field, see top 10 agent harnesses.

A safe way to use the credit

1. Pick one real task: a well-scoped issue in a repo you can afford to experiment on.
2. Write a spec file first, not just a prompt. Cloud sessions reward clear checklists because you are not watching. See our thoughts on planning without plan mode.
3. Review the diff like a PR. Treat output as a contribution from a contractor.
4. Track credit burn after the first run to estimate cost per task.
5. Decide before November 4 whether the workflow is worth paying for.

Prompts that suit unattended cloud work

A cloud session rewards prompts that would fail in a chat you were watching. Use structure.

• State the finish line. "Done means: all tests in packages/api pass, no new lint errors, and a PR description listing behavior changes."
• Bound the scope. "Only touch files under src/billing/. If you need to change anything else, stop and write the reason to NOTES.md."
• Ask for a running log. "Append what you tried and why to SESSION_LOG.md as you go." When you come back, the log is the fastest way to review.
• Require evidence. "Include the test output for each claim that something is fixed."
• Set a stop rule. "If two approaches fail, stop and summarize instead of continuing to try variations."

These rules turn a vague background job into something you can audit in five minutes.

Cost and risk questions to settle before you scale

1. What happens when the credit runs out? Find out how cloud usage is billed against your plan and whether you can cap it.
2. What can the session touch? It works through your GitHub connection. Review which repositories and permissions you granted, and prefer the narrowest scope.
3. Where are secrets? Do not put production credentials in a repo a cloud session can read. Use scoped test credentials.
4. Who reviews the output? Treat a cloud session's branch like a contributor's pull request, with the same review bar.
5. Does your company allow it? Sending source to hosted infrastructure may require approval, especially for regulated code.

Bottom line

General availability plus a $100 or $250 credit is a low-risk way to test whether laptop-closed agents change how you work. The credit deadline is October 7, the requirement is GitHub, and the catch is unknown pricing beyond the trial. If your workflow is GitHub-centered and you often hit local limits, claim it. If your code is local-only, watch Remote Control instead.

Details reflect Anthropic's September 24, 2026 announcement and may change. Terms apply; check the claim page for eligibility.

Related reading

• Claude Code Projects: one conversation coordinates parallel cloud threads
• Run Claude Code from your phone with Remote Control
• Ultraplan: cloud planning with browser review
• Will Claude Code kill plan mode?
• Claude Code pricing guide
• Cursor cloud agents on your own infrastructure
• Claude Code vs Cursor vs GitHub Copilot
• Official: ClaudeDevs announcement