Anthropic launches Remote Control to let developers supervise Claude Code agents from mobile devices

Anthropic has officially expanded the operational reach of its agentic coding tool, Claude Code, through a new feature called Remote Control.[1][2][3] This update marks a significant shift in how developers interact with artificial intelligence, moving away from the paradigm of being tethered to a high-powered workstation and toward a more flexible, supervision-based model. By enabling users to continue a locally running terminal session from a smartphone, tablet, or web browser, Anthropic is addressing a long-standing bottleneck in AI-assisted development: the inability to monitor and manage long-running agentic tasks while away from the primary machine. Since its initial research preview launch, Claude Code has rapidly become a cornerstone of Anthropic’s developer ecosystem, reportedly reaching a $2.5 billion annualized revenue run rate and capturing over half of the enterprise coding market share. This latest accessibility update underscores the company's ambition to transform the coding process from a manual labor-intensive task into an orchestrated workflow where the human acts as a mobile supervisor for an autonomous agent.

The technical implementation of Remote Control is designed to bridge the gap between local environment access and cloud-based convenience without sacrificing the developer's specific project configuration. Unlike standard web-based IDEs that run in an isolated cloud container, Remote Control acts as a secure synchronization layer for the existing Claude Code command-line interface. When a developer initiates a session using the new command or the internal shortcut, the tool generates a unique session URL and a QR code directly in the terminal. Scanning this code with the Claude mobile app on iOS or Android, or opening the URL in a browser, provides a real-time window into the local process.[1] Crucially, the code execution, filesystem modifications, and tool interactions continue to happen on the user’s local machine.[1][2][4] This ensures that all local Model Context Protocol servers, environment variables, and specialized build tools remain available to the agent regardless of where the human user is physically located.

This shift toward remote supervision is particularly relevant given the agentic nature of the tool. Unlike traditional autocomplete features, Claude Code is designed to handle complex, multi-step tasks such as refactoring entire modules, writing comprehensive test suites, or identifying and fixing bugs across dozens of files. These operations can take several minutes or even hours to complete. With the introduction of Remote Control, a developer can kick off a massive codebase migration at their desk, put their laptop in a bag, and monitor the agent’s progress while commuting or attending meetings. The interface allows for bidirectional interaction, meaning the user can approve file changes, answer clarifying questions from the agent, or provide new instructions directly from their mobile device. If the local machine enters a sleep state or the network connection drops, the system is designed to automatically reconnect once the machine is back online, with sessions timing out only after sustained outages of roughly ten minutes.[1][2]

Security remains a primary concern when allowing remote access to a local terminal, and Anthropic has implemented a specific architecture to mitigate these risks. The system does not require opening any inbound ports on the user’s firewall, which is a common vulnerability in traditional remote desktop setups. Instead, Claude Code makes outbound HTTPS requests to the Anthropic API, polling for work and streaming session data over a Transport Layer Security connection.[1] The authentication process utilizes multiple short-lived, narrowly scoped credentials that expire independently, limiting the potential impact if a single token is compromised.[1][4] However, the company has cautioned that the session URL itself functions as a temporary password, as anyone with access to the link could theoretically interact with the local session.[1] To further enhance security, users can toggle sandboxing flags during a remote session to enforce filesystem and network isolation, preventing the agent from accessing sensitive directories or making unauthorized external requests.

The competitive landscape for AI coding tools is intensifying, and Anthropic’s move to bring the CLI to mobile devices sets it apart from rivals like GitHub Copilot and Cursor. While GitHub Copilot has offered a mobile app for assigning tasks to its agent since mid-2025, that system runs entirely on GitHub’s cloud infrastructure via GitHub Actions, which may not have access to a developer's specific local setup.[1] Cursor, another popular AI-integrated editor, introduced a companion app earlier this year that relays prompts to a desktop IDE, but it lacks the deep terminal integration and agentic autonomy found in the CLI-first approach of Claude Code. Anthropic’s strategy appears to be paying off, as the tool has seen a reported 160 percent growth in active users over the last several months, fueled in part by the company’s acquisition of the Bun JavaScript runtime to optimize performance.

Currently, the Remote Control feature is being rolled out as a research preview specifically for users on the Max subscription tier, which costs between $100 and $200 per month.[1][2][3] Anthropic has indicated that access for Pro tier subscribers is imminent, though users on Team and Enterprise plans are currently excluded from the preview.[1][2] This exclusion is notable, as enterprise developers often maintain the most complex local environments that would benefit from remote monitoring.[1] The company has not yet provided a definitive timeline for when these organizational tiers will receive the update.[1] The release of Remote Control coincided with a broader corporate event where Anthropic unveiled its Cowork enterprise plugins, suggesting a bifurcated strategy: focused automation for large teams and high-mobility agentic power for individual power users.

As the AI industry moves toward more autonomous agents, the definition of a workstation is being rewritten. The ability to manage a professional-grade development environment from a smartphone suggests a future where the physical location of the programmer is increasingly decoupled from the compute environment. For Anthropic, Claude Code is no longer just a utility for the terminal; it is evolving into a persistent, omnipresent development partner.[5] This evolution reflects a broader trend in software engineering where the human role is shifting from "writing" code to "directing" it. By providing the tools to maintain that direction from any device, Anthropic is positioning itself at the center of this transition, betting that the most productive developers will be those who can manage their AI agents as easily as they check their email. The success of this model will likely depend on how well the company can balance this unprecedented flexibility with the rigorous security requirements of professional software development. For now, the introduction of Remote Control represents a significant milestone in making agentic AI a truly integrated part of the modern developer’s lifestyle, allowing for a level of continuity that was previously impossible in local development workflows.