Start a coding task in your terminal, close your laptop, and pick it up from a mobile device. That’s the promise behind Anthropic’s latest update to Claude Code.
The company recently introduced “Remote Control,” a feature that lets Claude Code users access and manage a local development session from a phone, tablet or browser. Claude continues running on the user’s own machine, while controls and outputs are mirrored through the mobile app or web interface.
Using Claude Code on the move
Claude Code, as you’ll know, is Anthropic’s terminal-based coding assistant, built to operate against a developer’s local files and environment.
Until now, sessions initiated in a terminal were largely tied to that machine and interface. The model itself runs in Anthropic’s cloud, but execution — file edits, shell commands and repository access — happens locally. And so Remote Control adds an access layer on top of that, relaying instructions and streams results back in real time.
To enable the feature, users run claude remote-controlclaude.ai/code. The interface mirrors the active terminal session, and commands, outputs and tool activity remain synchronized across devices, allowing a developer to move between their laptop and phone without restarting the task.
It’s worth noting that while Claude Code already has a web and mobile incarnation, those versions run in Anthropic’s hosted environment rather than against a live session on the developer’s own machine.
What you can (and can’t) do with Remote Control for Claude Code
Remote Control exposes the same local context available at the terminal. That includes access to the local filesystem, project configuration, connected tools and any configured MCP servers. Prompts sent from a phone or browser are processed against that live environment.
Conversations remain in sync across devices. A user can initiate a request in the terminal, check progress from their phone and issue follow-up instructions from the web interface, with the session maintaining continuity.
The feature also attempts to reconnect automatically if a device briefly loses connectivity or wakes from sleep, provided the underlying Claude process is still running.
However, there are some limitations. Each Claude Code session supports a single remote connection at a time. If the terminal process is closed or the claude command exits, the remote session ends and must be restarted manually.
Extended network disruption will also terminate the session. According to the documentation, if the machine remains online but cannot reach the network for roughly ten minutes, the process will exit.
So for now, Remote Control isn’t really about running unattended background agents for hours on end, but about giving developers mobility around an already active local session — checking progress from a meeting, reviewing output during a commute, or issuing a follow-up command away from their desk.
Community reacts: “A little janky”
Early feedback, while broadly positive in terms of concept, focused on the current limitations. Developer Simon Willison published some musings, describing the feature as “a little bit janky right now” after encountering errors, single-session limits and occasional API failures during testing.
Elsewhere, one user described difficulty interrupting running tasks, sessions that continue spinning after hitting stop, and intermittent UI disconnects. They also pointed to sessions failing to load, getting stuck in plan mode, exposing raw XML instead of interface elements, and limiting users to a single active connection. Navigating away from the Code view, they said, can cause the session to drop or take time to reappear.
“This is an extremely clunky and buggy prerelease, so don't try to hot fix prod from the toilet without a different mobile frontend,” they concluded.
Remote Control is available in research preview for subscribers on Anthropic’s Max plan, with Pro plan support expected later. It’s not clear when, or if, it will arrive on the Team or Enterprise plans.
