Anthropic Unleashes its AI Coder on the Web

Anthropic has launched its popular Claude Code AI assistant on the web and iOS, moving the powerful tool beyond the developer's terminal and into the browser for its Pro and Max subscribers. The move allows developers to delegate complex coding jobs to an autonomous AI agent that can write, test, and implement code on its own.

• Let Claude take the wheel: Users can now connect their GitHub repositories and assign tasks like bug fixes or feature updates that run in parallel on Anthropic’s cloud infrastructure. The agent works autonomously, provides real-time progress updates, and creates pull requests upon completion, effectively acting as a digital teammate.

• Putting up numbers: The launch pits Claude Code directly against rivals like Microsoft’s GitHub Copilot, but the tool is already a developer favorite. According to a report from TechCrunch, Claude Code has seen tenfold user growth since May and now accounts for over $500 million in annualized revenue.

• Safe in the sandbox: A heavy emphasis is being placed on security. Anthropic says every task runs in an "isolated sandbox environment" with strict controls on file and network access, and a secure proxy walls off the agent so it can only access pre-approved repositories.

The rise of agentic tools like Claude Code is accelerating a shift in the software development workflow, turning engineers from hands-on coders into managers who oversee and direct AI assistants. Anthropic is also open-sourcing its sandboxing feature for other developers to build on. While the increased autonomy is convenient, some argue it's a double-edged sword that makes rigorous code review more critical than ever. For developers who want to move between environments, sessions started on the web can also be transferred to the command-line to continue work locally.