How to Run Claude Code From Your Phone

Published July 21, 2026 · by the AQ team

You can run Claude Code from your phone in three ways as of July 2026: use Anthropic's built-in Remote Control feature to steer a session already running on a computer, build your own SSH and terminal-multiplexer stack to reach that same computer from any phone terminal app, or use a phone-native IDE like Lunel that turns the phone itself into the control surface. None of these run the agent's execution on the phone: in every case, the code, the filesystem, and the actual tool calls happen on a computer somewhere, and the phone is a window into that machine.

What "on your phone" actually means

Coding agents need a real filesystem, a shell, and often long-running processes (dev servers, test suites, builds). Phone operating systems do not give an app that kind of access to your project, so nobody, including Anthropic, runs the agent loop on the phone hardware itself. What differs between approaches is where that computer lives and how much of it you have to set up yourself:

• A vendor's official remote surface. The agent runs on your laptop or in the vendor's cloud, and the vendor's own mobile app or website gives you a chat-like view: approve permissions, read diffs, send follow-up prompts, get a push notification when it's done.
• Your own SSH stack. A generic terminal app connects to a server you control, where the CLI (Claude Code, Codex CLI, or anything else) runs inside a persistent terminal session.
• A phone-native IDE. An app built around mobile as the primary surface: file browser, git, and a full command line, with the agent as one feature among several.

Option 1: Anthropic's Remote Control

Claude Code's official mobile path is Remote Control, in research preview as of July 2026 and available on Pro, Max, Team, and Enterprise plans (not on API keys, and not when the endpoint is Bedrock, Google Cloud's Agent Platform, Microsoft Foundry, or a custom base URL). You start it from a project directory:

# Server mode: waits for remote connections, can host several sessions at once
claude remote-control

# Or add Remote Control to a normal interactive session
claude --remote-control

# Or, from inside an existing session
/remote-control

This prints a session URL and, with server mode, a QR code you scan with the Claude app (iOS and Android) to connect instantly. Claude Code keeps running on your machine the whole time: your filesystem, MCP servers, and project configuration never leave it, and only the conversation and tool-activity stream travel to the phone over an outbound HTTPS connection. The session reconnects automatically after your machine sleeps or your network drops, with one hard limit: if the machine can't reach the network for more than roughly 10 minutes, or you close the terminal or quit the process, the session ends. Push notifications can alert you when a long task finishes or needs a decision, and organizations can require Trusted Devices (biometric step-up, enrolled per device) before a phone can view or steer a session at all.

Other vendors ship the same pattern. OpenAI added Codex control to the ChatGPT mobile app on May 14, 2026: it pairs with Codex running on your Mac, laptop, or a remote box, and lets you review diffs, approve commands, and redirect tasks. Cursor's iOS app reached public beta on June 29, 2026 (no native Android app yet, though the cursor.com/agents web dashboard works in an Android browser); its cloud agents run in isolated VMs Cursor provisions, and a separate Remote Control mode lets you steer an agent already running on your own computer instead. GitHub shipped the same idea for Copilot's cloud agent: as of April 2026, GitHub Mobile can research a codebase, plan, make branch commits, and open a PR, and remote control for a session started in VS Code or the CLI is generally available on github.com and GitHub Mobile.

Option 2: Your own SSH and tmux stack

The vendor-neutral version of the same idea needs no special mobile feature at all: any CLI, on any server you control, reached from any terminal app.

# On the server (once)
tmux new -s agent
claude   # or codex, or any other CLI

# Detach: Ctrl-b then d. From your phone, later:
ssh dev-box
tmux attach -t agent

Three tools make this work well on a phone specifically. Mosh replaces the SSH transport with one that survives switching between Wi-Fi and cellular and tolerates high latency without dropping the session. Tailscale puts your server on a private mesh network so the phone can reach it without opening a public port or juggling a changing IP. And tmux (or screen) keeps the process alive independent of the connection, so a dropped signal never kills the agent mid-task. On iOS, Blink Shell and Moshi both build in mosh support; Termius is the more general cross-platform option (mosh on its free tier, tmux works as a plain command with no dedicated UI). On Android, Termux is a full Linux environment rather than an SSH client, so most people pair it with a client like Termius, or run the same server-side stack and connect with any mosh-capable app. The setup is the same one this site's guide to keeping Claude Code running after closing your laptop covers in more depth: the phone is just one more device attaching to a session that already lives elsewhere.

Option 3: A phone-native IDE (Lunel)

Lunel takes a different shape: instead of a chat view or a terminal emulator, it is a mobile app built as the primary interface, with the agent as one feature inside it. You run npx lunel-cli in a project on your existing machine, scan a QR code in the Lunel app, and it connects over an end-to-end encrypted channel. From there you get a file explorer, git operations, full command-line access to run scripts or start servers, and an AI agent that can run Claude Code, Codex, or OpenCode against the project, as of July 2026. Execution still happens on the machine running the CLI, not on the phone; Lunel's contribution is a purpose-built mobile UI around that connection rather than a terminal or a chat thread.

What actually breaks on a phone

Every approach above shares the same rough edges. Approving a permission prompt or reading a long diff on a small screen is slower than at a keyboard, so most people use the phone to check status and unblock an agent, not to drive fine-grained edits. Background execution is fragile on Android: aggressive battery optimization can delay push notifications or drop a backgrounded SSH connection unless you exempt the app. And exposing a terminal to the internet, even behind Tailscale or mosh, is still a real credential on a device you might lose, so the same device-loss plan you'd apply to a laptop applies here too.

Where AQ fits

AQ is the multiplayer coding harness where engineering teams run AI coding agents like Claude Code and Codex together: shared live terminals, a code editor, and app previews, in your own cloud. None of the setup above is necessary to check in from a phone, because the setup already happened when the workspace was created: agents run as real CLIs in persistent tmux sessions on your team's own VM, streamed live to the browser, so a session survives your laptop closing whether or not you ever open it again from a phone. Opening the same workspace URL in your phone's browser shows the same live terminal a teammate is watching on a desktop, with no SSH keys, QR-pairing step, or separate mobile app to install.

That browser view also carries the rest of the workspace: the live dev-server preview (shareable and guest links work from a phone browser too), and pinned comments on that preview, including screenshots, that get sent to the agent as prompts, useful for flagging something from a phone without touching a keyboard. Linear issues labeled ai-task turn into workspaces automatically, so a workspace can exist before you ever sit down at a laptop; a person still has to open it and start the agent. Each workspace gets its own isolated git worktree, agents commit and open PRs through your own GitHub login, and there is no shared multi-tenant execution tier: the VM is your team's.

AQ has two plans. Free is the full product, and you bring your own VM and pay only your cloud provider. Team is $100 per user per month (promotional pricing; standard is $200), billed monthly, and includes a dedicated, always-on VM that AQ manages in its own isolated network, with your rate locked for your first 12 months. If your team already runs agents on a shared box, checking in from a phone is just opening a browser tab, not standing up a new remote-access stack.