How to Run Claude Code From Your Phone

The reliable way to use Claude Code from a phone is to run it on a computer or server and treat the phone as a remote interface. That keeps the development environment intact and avoids trying to turn iOS or Android into a full workstation.

Here are five useful approaches, ordered from the simplest first-party experience to more configurable remote infrastructure.

Before choosing an interface #

Decide where the agent should run:

• Your laptop: no extra server, but remote access stops when it sleeps.
• A home server: always available if your network and power are reliable.
• A VPS: reachable and usually always on, but it needs patching, backups, and careful credential boundaries.

The host can read whatever the agent account can read. Avoid copying broad production credentials onto an always-on box just because the phone interface is convenient.

1. Anthropic Remote Control #

Claude Code’s Remote Control connects a running local session to Claude on the web or the iOS/Android app. Start it on the host:

claude remote-control

Open the session link or scan the QR code in the Claude app. Your repository, MCP servers, and local tools remain on the host.

Good for: the shortest supported route and a phone-friendly first-party interface.

Watch for: the Claude process and host must stay running, and the workflow is specific to Claude Code.

2. The official Telegram channel #

Claude Code Channels can connect an allowlisted Telegram bot to a running session. You can send tasks, receive replies, and handle supported tool permission requests with Allow and Deny buttons.

Good for: people who already use Telegram and want short conversational interactions.

Watch for: bot-token security, a host process that must stay alive, and a chat interface that is not a full terminal.

The complete installation and permission model are in How to Connect Claude Code to Telegram.

3. SSH from a mobile terminal #

A mobile SSH client connects directly to the development host. Add tmux so a network interruption does not kill the agent:

ssh dev@example-host
tmux new -As claude

Good for: full shell access, standard tooling, and easy recovery when a specialized interface fails.

Watch for: SSH key handling, host hardening, reachability, and the awkward ergonomics of a terminal UI on a small screen. Prefer a VPN or overlay network over a casually exposed public port.

4. A browser terminal #

Tools such as ttyd can expose a terminal through a browser. Put the terminal behind a private network or authenticated outbound tunnel and test that an unauthenticated request cannot reach it.

Good for: devices where a browser is easier than an SSH client.

Watch for: this is still remote shell access. Authentication, proxy headers, cookies, origin rules, updates, and session timeouts all become part of the security boundary.

5. Chatcode #

Chatcode connects an owned server to a browser terminal and Telegram using an outbound gateway. It can host Claude Code, Codex, Gemini, and OpenCode sessions through the same interface.

Good for: an always-on multi-agent workflow where browser and Telegram access are requirements.

Watch for: the server still needs administration, and Chatcode’s gateway and relay become additional dependencies to trust. SSH remains useful as a separate recovery path.

The step-by-step flow is in Claude Code in Telegram with Chatcode.

Which option fits? #

Start with the least infrastructure that meets the need. If Remote Control works, there is little value in building a tunnel. If you need an unrestricted shell, a chat bot will feel limiting no matter how polished it is.