How do you control Claude Code from WhatsApp?

Anthropic doesn't ship a WhatsApp integration for Claude Code: its Channels feature covers Telegram, Discord and iMessage, and a request for WhatsApp is still open. TextMyWorker fills that gap. It's a WhatsApp front end for the Claude Code you already have, running on your own computer. This guide walks through setup, a normal day, and what keeps it safe.

What you need

• A Mac or a Windows PC that can stay on and awake while you're away. Both are supported natively.
• A Claude plan. Claude Code runs on your own plan; if Claude Code isn't installed yet, setup installs it.
• WhatsApp on your phone. You don't need a Meta account, a bot token or a second number: your worker texts you from the TextMyWorker number.
• An invite. We're letting in a small group of design partners, free during the pilot.

No MCP. No plugin. Just WhatsApp. Nothing is installed into Claude Code's MCP setup, and we don't plug into your accounts.

Set it up in 3 steps

1. Ask the bot on WhatsApp. Message the TextMyWorker bot on WhatsApp (+972 55-315-3190), or use the button at the end of this page. It asks 3 short questions, including whether you use a Mac or Windows, and the founder approves you. It's a design-partner pilot, free during the pilot.
2. Run one command on your computer. On a Mac, paste this into Terminal: curl -fsSL https://relay.textmyworker.com/install | bash. On Windows, paste this into PowerShell: irm https://relay.textmyworker.com/install.ps1 | iex. A setup window opens. It installs Claude Code if it's missing, signs you in to Claude, lets you pick the folders your worker can use, and shows a QR code.
3. Scan the QR code and send the message. Scan the QR code with your phone. WhatsApp opens with a message already written. Send it, and your number is paired with that computer. There's no code to type.

The commands, ready to copy:

# Mac (Terminal)
curl -fsSL https://relay.textmyworker.com/install | bash

# Windows (PowerShell)
irm https://relay.textmyworker.com/install.ps1 | iex

Send your first task

Text it the way you'd text a colleague, or send a voice note. For example: "Run the tests in client-b and tell me what broke." The task runs in Claude Code on your computer, in the folder you chose, with your real files and tools. Voice notes are transcribed and quoted back, so you can see it heard you right.

Your message gets a reaction that changes as the work moves along:

• ⏳ waiting in line behind an earlier message
• 👀 received, starting
• ⚙️ working on it (for tasks that take more than about a minute)
• 🔐 waiting for your approval
• ✅ done
• ❌ failed

What a day looks like

• Morning. A summary arrives: what moved yesterday, what's waiting on you, and a suggested first step.
• On the way out. You send a voice note: "Fix the signup form bug and run the tests." Your message gets 👀, then ⚙️.
• At a client meeting. A Claude Code session wants to run git push. You get a 🔐 message with Allow and Deny buttons. You tap Allow and go back to the meeting.
• Picking up the kids. The long task finishes and you get the summary: what you asked, what happened, and how long it took.
• Evening. An evening summary of what got done and what's still open.

That's the idea behind "Claude Code To Go": left your laptop at home? Your work didn't stop. Dead time becomes shipping time.

The safety model

Your worker only gets the keys you hand it.

• Access levels. Start with one folder. Widen to several projects, and later to full access if you want it. Access can only be widened from your computer, never from WhatsApp, so someone holding your phone can't open up your computer. Narrowing works from anywhere.
• Risky actions wait for you. Deleting files, git push and history rewrites, deploys and publishing, sending email or messages, uploads and system commands send you a 🔐 Allow / Deny button first. Only a tap counts. Typing "yes" doesn't, and no answer within 5 minutes means no.
• "stop" stops it. Text stop and whatever is running stops.
• Only your number. Your computer only takes messages from the WhatsApp number it's paired with.
• Its own number, on the official API. The worker texts you from the TextMyWorker business number on Meta's WhatsApp Cloud API. Your personal WhatsApp is never linked to a bot.
• A relay that passes messages, not keeps them. The relay holds message content in memory until your computer picks it up and never writes it to disk. Your files stay on your computer; the parts Claude reads for a task go to Anthropic, as with any Claude Code use.
• A local log. Every action, and whether it ran, was approved or was blocked, is logged on your computer and never sent to us.

Honest limits: on a Mac, the one-folder level also runs terminal commands in Claude Code's macOS sandbox. Windows has no system sandbox, so there the folder limit applies to Claude's file tools and the approval buttons do the rest. Full access has no folder fence on either system. The full details are in How it works & your data.

Limits worth knowing

• The computer has to be on and awake for work to run. While it's plugged in, TextMyWorker keeps it awake and reconnects after sleep. A closed laptop without an external display will still sleep.
• WhatsApp allows free-form messages only within 24 hours of your last message. Outside that window, a short approved template goes out first.
• Linux isn't supported today.

Other ways to reach Claude Code from a chat app

If WhatsApp isn't your app, Anthropic's Channels plugins connect a running Claude Code session to Telegram, Discord or iMessage. Community projects also bridge WhatsApp by linking your account through unofficial libraries; their authors warn about bans and suggest a separate number. All the options are compared in How to use Claude Code from your phone.

FAQ