Anthropic’s AI-powered coding agent, Claude Code, is rapidly boosting the company’s internal productivity and has already generated over $500 million in annualized revenue, the company said as it expanded the tool’s availability to web and mobile platforms.
The artificial intelligence firm reported a 67% increase in merges per engineer per day, even as its engineering organization doubled in size. Anthropic also stated that 90% of the Claude Code tool itself was written using Claude Code.
The expanded access allows developers with paid “Pro” and “Max” accounts to run Claude Code through Claude.ai’s web interface and mobile application. This move shifts the tool from a command-line interface towards more visual and accessible platforms.
Claude Code is designed to automate programming tasks, including code reviews and modifications to repositories. It enables agents to clone repositories, execute tests, and apply changes, thereby freeing human engineers from repetitive work.
Developers can now initiate coding workflows directly from the web or mobile app, with tasks executed on instances managed by Anthropic. A central feature is its integration with GitHub, allowing users to connect repositories and assign specific tasks.
The new interfaces display the agent’s progress and reasoning in real time, offering greater transparency. Developers can also supervise and guide the agent while it works, adjusting priorities or providing new instructions without stopping current executions.
The mobile capability means developers can start a task on their phone and later review the results, with Claude Code continuing the work in the background. This aims to increase efficiency by enabling asynchronous work and parallel task processing.
Anthropic emphasized security measures, stating that each coding session runs in an isolated sandbox environment. All interactions with Git repositories are routed through a secure proxy service to control access.
Users can configure network access, allowing connections to trusted networks for package downloads or restricting access entirely. This flexibility enables teams to balance productivity with strict security requirements.
The expansion into web and mobile is a notable contrast to some competitors, such as Google’s Jules, which moved from a web interface to a command-line approach. Anthropic also offers Claude Code GitHub Actions, allowing agents to work within a user’s GitHub Actions environment.
These options provide flexibility for teams to choose execution environments based on their security policies and operational control needs. The growing trend of AI coding agents underscores a shift toward more automated development toolkits.