Claude Code remote control: the built-in option and the local-only one

If you want to control Claude Code from your phone, the first thing to know is that Anthropic now ships this themselves: a first-party feature literally called Remote Control, included with a Claude subscription at no extra cost. It's good, and for a lot of people it's the whole answer. This page explains exactly what it does and how it's built — then where our own tool, DevFob, takes a different architectural path. Use whichever matches how you work.

The built-in: Anthropic's Remote Control

Remote Control (currently a research preview) connects a Claude Code session running on your machine to claude.ai/code in any browser, or to the official Claude mobile app on iOS and Android. The session itself keeps running locally — your filesystem, MCP servers, and project config stay on your machine — and the phone or browser is a synced window into it.

Getting started is about as low-friction as it gets:

• Start it with one command. Run claude remote-control for a dedicated server mode (it can host multiple concurrent sessions, with an optional git-worktree-per-session mode), claude --remote-control for a normal interactive session that's also reachable remotely, or type /remote-control inside a session you already have going — it carries your conversation history over. There's a VS Code path too.
• Connect by QR code or URL. The terminal shows a session URL and can display a QR code that opens the session straight in the Claude app. Sessions also just appear in your session list at claude.ai/code and in the app.
• Full conversation from the phone. You can type new prompts, watch subagent and workflow progress, send images and files from your phone into the local session, and even switch models — it's genuinely the session, not a notification feed.
• Push notifications. With the Claude app installed, Claude can push to your phone when a long task finishes or when it needs a decision; there are toggles in /config for proactive pushes and action-required pushes.

Requirements, per the official docs: a claude.ai subscription login (Pro, Max, Team, or Enterprise — API-key auth isn't supported), and on Team/Enterprise an Owner has to switch it on in admin settings first. It's not available on Amazon Bedrock, Google Cloud's Agent Platform, or Microsoft Foundry, nor in gateway/proxy setups that point ANTHROPIC_BASE_URL somewhere other than Anthropic's API.

How it's built: relayed through Anthropic's cloud

The architecture is clever and worth understanding, because it's the root of both the feature's biggest strength and its main trade-off. Your machine makes outbound HTTPS requests only — no inbound ports opened, nothing exposed on your network. Claude Code registers with the Anthropic API, and when you connect from a phone or browser, Anthropic's servers route messages between the remote client and your local session over a streaming connection, secured with TLS and short-lived single-purpose credentials.

Two consequences follow directly from that design:

• It works from anywhere. Coffee shop, train, another country — if your phone has internet and your machine has internet, they can find each other, because Anthropic's infrastructure is the meeting point. No pairing, no same-network requirement.
• Your session transcript lives on Anthropic's servers while connected. The docs are commendably clear about this: while Remote Control is active, the transcript — your messages, Claude's responses, and tool activity — is stored on Anthropic's servers to keep devices in sync and survive reconnects, and is retained under their data-usage policy. Execution and file access stay local; the conversation transits and rests in their cloud. The docs also note that organizations with compliance requirements such as Zero Data Retention can't enable the feature.

On session lifetime: Remote Control is a local process, so it lives and dies with that process. Close the terminal and the session ends. If your laptop sleeps or the network drops, the docs say the session “reconnects automatically when your machine comes back online” — which in practice means a lid-closed MacBook on battery isn't running your session in the meantime; the work resumes when the machine wakes. And if the machine is awake but offline for more than roughly 10 minutes, the session times out and the process exits.

The local-only path: DevFob

Disclosure: DevFob is our product — this is our site, so read this section with that in mind. DevFob is a Mac menu-bar app ($15 once, 14-day trial, macOS 14+) with free iPhone and Apple Watch companion apps (Android and Wear OS companions are in beta). It solves an overlapping problem with a different architecture, and the differences are exactly where the two tools diverge in philosophy:

• Nothing leaves your devices. Your Mac and phone talk directly over your LAN or a personal hotspot, end-to-end encrypted (Curve25519 pairing, ChaCha20-Poly1305). There's no cloud relay, no server-side transcript, no account to create, and zero analytics in the app. The trade-off is the mirror image of the built-in's: your phone needs to reach your Mac directly, so "from anywhere on the internet" isn't the design goal — your own network and hotspot range are.
• It watches Codex too. DevFob mirrors both Claude Code and OpenAI's Codex CLI sessions in one place. Anthropic's Remote Control is, understandably, Claude-only.
• Structured interaction, not a chat window. Instead of a full conversation view, DevFob surfaces the moments that actually block an agent: multi-choice questions as tappable forms, plan approvals, and permission gates — designed to be answered from a phone or, genuinely, an Apple Watch. If you want to type long new prompts remotely, the built-in is better at that.
• Per-account rate-limit meters and auto-switch. Live gauges for the 5-hour and weekly usage windows for every Claude account you keep, with optional auto-switch to the login with the most headroom when a window fills.
• Notch HUD. A glanceable status readout in the Mac's notch area, so you see what every session is doing without switching windows.
• Nomad mode. Keeps sessions running with the MacBook lid closed — the carry-it-across-the-house scenario where the built-in's session would wait for the machine to wake.
• Any terminal, existing sessions. DevFob observes the Claude Code sessions you already run in whatever terminal you use — no special launch command or server mode.

Side by side

When the built-in is the right answer

We build on Anthropic's platform and we'd rather you pick the right tool than our tool. Use Remote Control — and skip DevFob — when:

• You want to steer sessions from anywhere — a different city, mobile data, a borrowed browser — with no pairing and no same-network constraint.
• You want to type full prompts from your phone, attach screenshots, and treat the phone as a real second seat for the conversation.
• Zero extra software and zero extra cost matter. It's one command and the Claude app you may already have.
• You only run Claude Code, on one account, and don't need Codex coverage or per-account usage meters.
• Cloud-relayed transcripts are fine under your policies. For most individual developers they are; the docs spell out exactly what's stored.
• You want first-party support that evolves with Claude Code itself — subagent progress, workflows, and new capabilities land there first.

DevFob earns its $15 in the complement of that list: when the transcript must stay off third-party servers, when you also run Codex, when you want your wrist to buzz for a plan approval, when you juggle multiple accounts against rate-limit windows, or when the MacBook needs to keep working with the lid shut. Plenty of people will reasonably use both — the built-in for far-from-home steering, DevFob for everything within Wi-Fi range.

Accurate to the best of our knowledge as of July 2026, based on Anthropic's official Remote Control documentation. Remote Control is a research preview and changes quickly; the linked docs are the source of truth. Something wrong here? Tell us.