Earlier, we wrote about Cursor giving AI coding agents their own cloud computers. Today, Anthropic is going the opposite direction. Claude Code Remote Control, announced on February 24, lets developers start a coding session in their terminal and continue it on a phone, tablet, or any browser — while everything runs on their local machine.
No code moves to the cloud. No environment gets replicated somewhere else. Your phone is just a window into the session still running on your computer, with full access to your filesystem, MCP servers, tools, and project configuration.
The distinction matters. In a market where AI coding tools are racing toward cloud execution, Anthropic is betting that many developers want their agent close to home — they just don’t want to be chained to their desk while it works.
How it Works
Run claude remote-control in your project directory. The process stays running, registers with the Anthropic API, and waits for connections. You get a session URL and a QR code. Open the URL in any browser, scan the code with the Claude mobile app, or find the session in your session list on claude.ai/code.
The connection uses outbound HTTPS only — no inbound ports open on your machine. All traffic routes through the Anthropic API over TLS using short-lived credentials, each scoped to a single purpose and expiring independently. When the session ends, the remote link becomes invalid.
If your laptop sleeps or your network drops, the session reconnects automatically when your machine comes back online. If the network stays down for more than roughly 10 minutes, the session times out.
The conversation stays in sync across all connected devices — send messages from your terminal, browser, and phone interchangeably. Each Claude Code instance supports one remote session at a time.
Why Local-First Matters
Before Remote Control, developers who wanted mobile access to Claude Code sessions built workarounds: SSH tunnels, tmux sessions, ngrok proxies, custom WebSocket bridges. Functional but fragile.
Remote Control replaces all of that with a native streaming connection. But the real story is the architectural choice behind it.
Cursor’s cloud agents, which we covered yesterday, move the entire execution environment to the cloud. Each agent gets its own VM with a full development environment. That model works well for parallelizing tasks and eliminating local resource conflicts. But it means your code runs on someone else’s infrastructure.
Claude Code Remote Control takes the opposite approach. Your code, your environment, your MCP servers — everything stays on your machine. The cloud handles routing messages between your devices and your local session. Nothing else.
For teams with strict data residency requirements, proprietary codebases, or custom toolchains that don’t translate well to cloud environments, that distinction is significant. Your security posture doesn’t change when you pick up your phone.
According to Mitch Ashley, VP and practice lead for software lifecycle engineering at The Futurum Group, “Anthropic’s local-first architecture in Claude Code Remote Control shows us how quickly innovation loops are collapsing. Just as DevOps loops enabled smaller software releases at higher velocities, AI is shortening the shelf life of the prized first-mover advantage.”
The Business Context
Claude Code has reached $2.5 billion in annualized run rate as of February 2026 — more than doubled since the start of the year — with 29 million daily installs in Visual Studio Code.
Remote Control is currently a research preview on Pro and Max plans. It’s not yet available on Team or Enterprise plans — a notable gap given that enterprises with security-sensitive codebases would benefit most from the local execution model.
The feature shipped the same week as Cursor’s cloud agents, Claude Code Security, and Anthropic’s COBOL modernization announcement. AI coding tools are moving from “useful assistant” to “core development infrastructure” faster than most teams planned for.
What This Means for DevOps Teams
Remote Control solves a specific workflow problem: the long-running task you can’t babysit. Start a refactoring job, a test suite, or a deployment pipeline in Claude Code. Walk away. Monitor and steer from your phone. Come back when it’s done.
That’s a practical improvement, but the broader implication is about agent supervision patterns. As AI coding agents take on more autonomous work — whether running locally or in the cloud — the interface between developer and agent matters more. Remote Control makes supervision location-independent. You don’t need to sit at your terminal to keep an eye on what the agent is doing.
The one-session-at-a-time limitation is worth noting. Unlike Cursor’s approach, where you can run 10 to 20 cloud agents in parallel, Remote Control is a single-thread model. You direct one agent from anywhere, not many agents from one place. For teams that need parallel execution, the cloud model still wins. For teams that need local control with mobile flexibility, this fills the gap.
The two approaches aren’t mutually exclusive. Developers will likely use both cloud agents for parallelized, well-defined tasks and local agents with remote access for work that requires their specific environment, tools, and data.
Remote Control is available now. Run claude remote-control or /rc in any Claude Code session on a Pro or Max plan.
