Claude Chat Mobile CLI
Claude Chat Mobile opens the same Claude Code session on your computer from a phone: approvals, questions and lock-screen pushes.
- 55 ★ on the repository
- Last commit
- Apache-2.0
Broad access
Why this level
- It opens control of your Claude Code from a phone: the token and address give access to the computer.
- Tool approvals can be granted remotely, including always allow.
Install
- Works in
Manual install
curl -fsSL https://github.com/Ike-li/claude-chat-mobile/archive/refs/heads/master.tar.gz | tar xzA step from the README three-step section; then `cd claude-chat-mobile-master` and `npm ci --omit=dev && npm run setup && npm start`. Review the archive before unpacking.
Access and checks
How we reviewBroad accessThe tool can change data, run commands or handle secrets.
- It opens control of your Claude Code from a phone: the token and address give access to the computer.
- Tool approvals can be granted remotely, including always allow.
This is third-party code. Review the repository files before installing.
What it does
Claude Chat Mobile is a self-hosted web chat that drives the real local `claude` through the Agent SDK. The same CLAUDE.md, MCP, skills, login and permission rules apply on the phone. From the screen you can allow or deny a tool call, answer an agent question, interrupt a turn, change model and permission level, review changes and rewind a session. It works with third-party gateways, API keys, Bedrock and Vertex, which official Remote Control does not cover. The service, session records, device trust and audit stay on your computer. It installs as a PWA, and a macOS desktop console exists.
Who it is for. People who run long Claude Code tasks and do not want the agent to wait for an approval overnight while the owner is away from the computer.
Good fit when
- An agent stopped on an approval prompt while you are away from home.
- Claude Code runs through a third-party gateway or an API key, so official Remote Control is unavailable.
- You want to see changes and background task status from a phone.
- You need to continue the same session you started in the terminal.
Not a fit when
- Official Remote Control with a claude.ai login is enough.
- Several devices must write to one session at the same time.
- Remote access to the computer is forbidden even through Tailscale.
Example request
Show the Claude Code approval prompt on the phone and allow file reads in the project only.Limitations
Node.js 20 or newer and a working `claude` are required, because the project does not install or sign in to Claude. A single driver controls a session, so terminal and phone do not write at once. Public access needs protection: `AUTH_TOKEN` is mandatory and a new device needs approval, yet the token, address and `doctor` output must never be posted. The README is in Chinese, with an English version.
How to disable. Stop the server and delete the project folder. Then remove trusted devices with `node scripts/device.js` and close the tunnel or Tailscale access.
FAQ
How do I start it?
Download the repository archive, run `npm ci --omit=dev && npm run setup && npm start` and open the printed address on the phone.
How do I approve a new device?
Run `node scripts/device.js list`, then `approve <ID>`. A QR code from `node scripts/qr.js` is easier than typing the token.
How is it different from official Remote Control?
It works with API keys, third-party gateways, Bedrock and Vertex, keeps records with you and shows every session of a folder. It does lack multi-device co-driving.
What can I do from the phone?
Allow or deny a tool, answer a question, interrupt a turn, change model, view a diff and rewind a session with `/rewind`.
Are there notifications?
Yes, lock-screen pushes arrive for approval requests, agent questions and finished background tasks.
Discussion
No questions or reviews yet.
Related
CLI toolsA Telegram bot for remote access to Claude Code: chat and the agent edits your project, with sessions and directory sandboxing.
A pixel-art office on screen shows Claude Code agents at work: who is busy, who awaits approval and what each session costs.
Streams Claude Code hidden thinking, tool calls and subagents to a separate terminal with a live session tree
An MCP server that gives Claude Desktop and other clients file tools, a shell and sub-agents in the style of Claude Code
Rust terminal client claude-rs for Claude Code: the session stays in terminal scrollback, output does not flicker, tmux works.
Reads Claude Code session logs as a conversation: tool calls, token counts and a live tail of a running session.
Comparison with similar tools
Cards that share the main search query, and how they differ.
| Tool | Official | Repository stars | Main difference |
|---|---|---|---|
| Claude Chat Mobile (this card) | no | 55 | |
| MobileCode | no | 244 | a mobile client for coding agents; Claude Chat Mobile drives your real `claude` with its rules, skills and MCP |
| claude-session-driver | no | 114 | drives Claude Code sessions from other processes; Claude Chat Mobile adds a phone web interface |
README in short
The app drives your own `claude` from the inside instead of replacing it: rules, MCP and skills stay the same. Model requests go out as usual, from the CLI.