How to Build and Deploy Software from Your Phone with Claude Code Remote Control

Six months ago, if you told me I'd be pushing production code from my couch (phone in one hand, coffee in the other) I would have laughed.

Now it's Tuesday.

Claude Code recently shipped a set of features that quietly changed how I build software for clients: Remote Control, Channels, and server mode. Together, they turned my phone into a real development environment. Not a demo. Real client projects, real deployments, real revenue.

If you're new to Claude Code, our complete guide to Claude Code in 2026 covers the fundamentals. This article picks up where that leaves off.

Here's what happened, how it works, and how you can set it up yourself.

I build websites and web applications for clients. Law firms, service businesses, local companies. Every project used to mean sitting at my desk, terminal open, juggling files and deployments.

Client calls during lunch? Close the laptop. Pick it up later. Lose context. Re-read where I left off. Waste 20 minutes getting back into flow.

The bottleneck was never the code. It was me needing to be physically in front of my machine.

Three features removed that bottleneck entirely.

Remote Control lets you keep a Claude Code session running on your local machine, with full access to your project files, tools, Git, MCP servers, everything, while driving it from any browser or the Claude mobile app.

You run one command:

claude --remote-control "Client Project"

The session appears automatically in the Claude app on your phone. Tap the "Code" button and it's there. You also get a direct session URL you can open in any browser. You're now controlling your full local development environment from your pocket.

Your session stays on your machine under the hood. Nothing moves to the cloud. Anthropic's servers route messages between your phone and your local session over a TLS-encrypted connection. No inbound ports open on your machine. If your laptop sleeps or your network drops, the session auto-reconnects when you're back online.

This isn't a watered-down mobile editor. It's your actual filesystem, your actual config, your actual deployment pipeline. Accessible through your phone screen.

You can also activate it mid-conversation if you're already in a session (the Remote Control docs cover all the options):

/remote-control

Or enable it globally for all sessions through /config.

Channels push external events directly into your running Claude Code session. Think of them as bridges between the outside world and your terminal.

Set up a Telegram or Discord bot, connect it to Claude Code, and you can text commands to your development environment from anywhere:

• "What's the build status?"
• "Fix the header spacing on the homepage."
• "Push to main."

Claude receives the message, acts on it, and replies. All within your Telegram or Discord chat.

The setup involves creating a Telegram bot, installing the channel plugin in Claude Code, and pairing your account. The full step-by-step is in the setup section below, and it takes about 10 minutes from scratch.

Once it's running, every message you send to your bot lands in your Claude Code session. Claude can read your project files, run commands, make changes, and reply back to you in Telegram. Only your approved account can trigger events. Everyone else gets silently dropped.

Discord works the same way. The full setup for both platforms is also in the Channels documentation.

Server mode turns your machine into a persistent session server. Multiple concurrent sessions, each isolated in its own Git worktree, running in the background even when you're not connected.

The key difference from a regular --remote-control session: server mode uses the subcommand claude remote-control (no -- flag). A regular remote session is a single interactive session. Server mode manages multiple sessions and keeps them alive independently.

You can also combine server mode with Channels, giving you both phone-based session control and Telegram messaging into the same server. The full setup is in Step 5 below.

I've used this to run work across multiple client projects at the same time, reviewing one site's homepage while another session handles a different client's schema markup.

Let me walk through two real projects from last week. I'm keeping details light (no client names or proprietary specifics) but the workflow was real.

Project 1: A service business website

A laundry pickup service needed a full website. SEO-optimized pages for each service area, schema markup, booking integration. The kind of project that normally chains me to my desk for a full day. We've done similar work before — our Freshly Folded case study covers how we approach these builds.

I started the session at my desk in the morning. Set up the project scaffold, pointed Claude at the design tokens and content brief. Then I left for a dentist appointment.

From the waiting room, I opened the session on my phone through Remote Control. Reviewed the homepage copy. Asked for adjustments to the CTA placement. Approved the schema markup. Moved through the service pages one by one.

By the time I got home, the site was 80% done. I sat down, reviewed the final build, and told Claude to push to GitHub. Vercel picked up the commit and deployed it. Production in under a minute.

Total desk time for a full website: maybe two hours.

Project 2: A law firm site

A law firm needed their site rebuilt. New practice area pages, attorney bios, case results, structured data. Heavy on local SEO and JSON-LD schema.

I kicked off the project at my desk, then shifted to the couch with my phone. Through Remote Control, I reviewed each page as Claude built it. Tweaked the consultation CTA. Adjusted the FAQ sections. Checked the schema was wired correctly.

When everything looked right, I told Claude to commit and push. GitHub Actions ran the build. The site went live.

I didn't open my laptop once during the entire review and deployment phase.

Not the building. The deploying. That's what changed the most.

My workflow now looks like this:

1. Claude writes the code in my local environment
2. I review changes from my phone via Remote Control
3. I say "commit and push to main"
4. Claude commits with a proper message, pushes to GitHub
5. CI/CD picks it up (Vercel, Netlify, whatever the project uses)
6. Production deployment happens automatically
7. Claude confirms the push went through

If you want to understand how Claude Code handles Git operations, CI/CD pipelines, and deployment automation in depth, our article on Claude Code security and enterprise workflows walks through the full picture.

From idea to production. From my phone. No SSH, no laptop, no context switching.

For hotfixes, it's even faster. A client reports a bug via Telegram. The message hits my channel. Claude reads it, investigates the codebase, finds the issue, fixes it, and pushes. Sometimes before I even read the original message. I just get a Telegram reply: "Fixed. The contact form wasn't validating email addresses correctly. Pushed to main, deploying now."

That loop, bug report to production fix, can happen in minutes, without me touching a keyboard.

If you want to try this yourself, here's the full setup. If you don't have Claude Code installed yet, our development environment setup guide covers everything from Node.js to your first session.

Prerequisites

• Claude Code v2.1.51 or later (check with claude --version)
• A Claude Pro, Max, Team, or Enterprise plan. Remote Control also works on eligible claude.ai accounts, but API keys are not supported.

Step 1: Install or update Claude Code

curl -fsSL https://claude.ai/install.sh | bash

If you are on Windows PowerShell, use irm https://claude.ai/install.ps1 | iex instead.

Step 2: Sign in

claude
/login

Step 3: Start your first remote session

claude --remote-control

The session shows up automatically in the Claude app on your phone. Tap "Code" and you're in. You can also open the session URL in any browser.

Step 4 (optional): Add Telegram channels

First, install Bun if you don't have it (required for channel plugins):

curl -fsSL https://bun.sh/install | bash

Restart your terminal after installing.

Next, create your bot on Telegram. Open Telegram, search for @BotFather, and send /newbot. Pick a name and username (must end in "bot"). Copy the API token BotFather gives you.

Now open Claude Code and install the plugin:

/plugin marketplace add anthropics/claude-plugins-official
/plugin install telegram@claude-plugins-official

Configure your bot token:

/telegram:configure

Paste your BotFather API token when prompted.

Exit Claude Code and restart with the channel enabled:

claude --channels plugin:telegram@claude-plugins-official

Finally, pair your account. Send any message to your bot on Telegram. You'll receive a pairing code. Back in Claude Code:

/telegram:access pair <the-code-you-received>
/telegram:access policy allowlist

The first command pairs your Telegram account. The second locks access so only you can send commands.

Instead of a single remote session, you can run your machine as a server that manages multiple sessions in the background. Each session gets its own isolated Git worktree, which is the same concept covered in our guide to agent teams and parallel development.

claude remote-control --name "My Projects" --spawn worktree --capacity 4

Breaking this down:

• remote-control (no -- flag) starts server mode instead of a single session
• --name "My Projects" labels the server so you can find it on your phone
• --spawn worktree gives each session its own isolated Git worktree (separate branch, no conflicts)
• --capacity 4 allows up to 4 concurrent sessions

The server keeps running even when you disconnect. Open the Claude app on your phone, tap "Code," and select your named server to connect. Each connection gets its own worktree. Disconnect and reconnect whenever you want.

To add Telegram channels to server mode:

claude remote-control --name "My Projects" --spawn worktree --channels plugin:telegram@claude-plugins-official

I'm not going to pretend this replaces a full desk setup for all software development. Complex systems architecture, deep debugging sessions, large-scale refactors. You still want a proper screen and keyboard for those.

But if any of this sounds like you, this workflow will change your output:

• Freelancers and agency developers juggling multiple client projects
• Solo founders who need to ship while living their life
• DevOps engineers who get paged while away from their desk
• Anyone whose development work includes building pages, writing features, fixing bugs, or deploying updates

If you build full-stack applications with Claude Code, Remote Control means you can review and deploy that work from anywhere.

For the 70% of dev work that fits those categories, the desk is now optional.

I shipped more client work last month than any month before. Not because I worked more hours. Because dead time became productive time. Waiting rooms, commutes, lunch breaks. Five minutes here, ten minutes there. It adds up fast.

The barrier to building and deploying software dropped from "sit at a computer with a full development environment" to "have a phone and an internet connection."

That's not an incremental improvement. That's a different game entirely.

---

I build websites and web applications for service businesses and law firms at Luminous Digital Visions. If you want to see what this workflow can do for your next project, book a free 30-minute call.