Anthropic Launches Claude Code Cloud Sessions

Anthropic has transitioned its Claude Code cloud sessions from research preview to general availability. This release allows developers to execute complex, long-running programming tasks—such as system migrations, extensive refactoring, dependency updates, and automated test loops—directly on Anthropic-hosted infrastructure. Because these sessions run remotely, developers can initiate a task, close their laptops, and return later to review the results without needing to keep their local machines awake and connected.

To encourage adoption, Anthropic is offering promotional credits to existing subscribers, which must be claimed by October 7 using the command /claim-credit. Subscribers on the Pro plan receive a one-time $100 credit, while Max plan users get $250. These credits apply automatically to cloud sessions and operate outside of standard plan usage limits, meaning tasks can continue running even after a user's regular monthly allowance is exhausted.

Developers can launch these remote sessions through multiple interfaces, including the web at claude.ai/code, the desktop app, the mobile app, or by running claude --cloud in the command line. A connected GitHub account is required, as each session automatically checks out a copy of the repository and operates on a separate branch. This setup ensures that the AI agent's output goes through standard review paths, such as pull requests and continuous integration checks, before merging.

For organizations with strict security or data residency requirements, Anthropic is offering a self-hosted option in public beta. This allows enterprise teams to route sessions to runners inside their own networks using partner platforms like Coder. This deployment model ensures that repository checkouts, build artifacts, and secrets remain entirely on organization-controlled infrastructure, positioning Claude Code to compete directly with asynchronous coding tools like GitHub Copilot Workspace, Cognition Devin, and Google Jules.