Control AI Agents from Your Phone
Getting started with LeCoder MConnect for mobile AI development workflows.
You're running Claude Code on your laptop. It's working through a complex refactoring task. You need to step away—grab coffee, take a meeting, go for a walk. But you want to monitor progress. Maybe send a quick instruction if it gets stuck.
This is exactly what LeCoder MConnect does.
The Problem with Current AI Agent Workflows
Today's AI coding assistants are powerful but tethered:
- Claude Code runs in your terminal
- Gemini CLI needs an active session
- Cursor requires the app open
- Aider is command-line only
If you close your laptop, the session dies. If you're on your phone, you can't interact. You're stuck at your desk watching text scroll.
The Solution: MConnect
MConnect creates a secure bridge between your AI agents and any device with a browser.
Your Machine (AI Agents) ←→ Cloudflare Tunnel ←→ Your Phone (Browser)
No port forwarding. No VPN. No accounts. Just a QR code.
Quick Start
1. Install Prerequisites
# macOS
brew install cloudflared
# Ubuntu/Debian
curl -L https://github.com/cloudflare/cloudflared/releases/latest/download/cloudflared-linux-amd64.deb -o cloudflared.deb
sudo dpkg -i cloudflared.deb
2. Run MConnect
npx lecoder-mconnect
You'll see:
╔═══════════════════════════════════════╗
║ LeCoder MConnect v0.1.4 ║
╠═══════════════════════════════════════╣
║ Local: http://localhost:3456 ║
║ Remote: https://abc123.trycloudflare.com ║
╠═══════════════════════════════════════╣
║ Scan QR code to connect: ║
║ ║
║ [QR CODE APPEARS HERE] ║
║ ║
╚═══════════════════════════════════════╝
3. Scan from Your Phone
Point your phone camera at the QR code. It opens the terminal interface in your mobile browser.
4. Start an AI Agent
Now start your AI agent in the terminal:
claude
Watch it appear on your phone in real-time.
Real-World Workflows
Workflow 1: Monitor Long Tasks
You've asked Claude to refactor a large codebase. It'll take 30 minutes.
# Start MConnect with Claude
npx lecoder-mconnect --preset single
Scan the QR, then walk away. Check progress from your phone whenever you want. If Claude asks a question, type your response from your phone.
Workflow 2: Parallel Agents
You want Claude for implementation and Gemini for research, running simultaneously.
npx lecoder-mconnect --preset custom --agents '[
{"type":"claude","name":"Implement"},
{"type":"gemini","name":"Research"}
]'
On your phone, you'll see two terminal panes. Switch between them with a tap.
Workflow 3: Safe Monitoring
You want to watch but not accidentally send commands.
npx lecoder-mconnect --read-only
Read-only mode by default. You can see everything but can't type until you explicitly enable input.
Workflow 4: Team Sharing
Share the tunnel URL with a teammate. They can watch the same session from their device.
https://abc123.trycloudflare.com
Coming soon: Real-time cursors and chat alongside the terminal.
Mobile UI Features
The MConnect mobile interface is designed for touch:
- Smooth scrolling: Native-feeling scroll through terminal history
- Pinch to zoom: Make text larger or smaller
- Tap to focus: Select which agent to interact with
- Pull to refresh: Reconnect if connection drops
- Landscape mode: Full-width terminal on tablets
Security
Every feature was designed with security in mind:
| Feature | How It Works |
| No accounts | No signup, no login, no data stored |
| Encrypted tunnel | All traffic through Cloudflare TLS |
| Ephemeral URLs | New URL each time, expires when you stop |
| Token auth | Unique token per session |
| Read-only default | Must explicitly enable input |
| Guardrails | Block dangerous commands |
Common Use Cases
Developer on the Go
Run AI agents on your desktop while commuting. Review and guide from your phone.
Pair Programming
Share your tunnel URL with a colleague. They can watch and suggest improvements.
Long-Running Tasks
Start a complex generation, go to a meeting, check progress during breaks.
Home Lab Setup
Run MConnect on a Raspberry Pi. Control from your iPad anywhere in the house.
Demo and Teaching
Share your AI workflow with students or clients through the tunnel URL.
Troubleshooting
QR code won't scan
- Ensure good lighting
- Try zooming in on the terminal
- Use the URL directly instead
Connection drops
- Check your internet connection
- Cloudflare tunnels reconnect automatically
- If persistent, restart MConnect
Slow on mobile
- MConnect streams efficiently but large outputs take bandwidth
- On slow connections, use read-only mode to reduce data
Agent not starting
- Ensure the AI tool is installed (
claude,gemini, etc.) - Check you have valid API keys configured
- Try
--preset shell-onlyfirst
What's Coming
We're actively building:
- Container isolation: Each agent in its own sandbox
- Git worktrees: Parallel branches per agent
- Ollama support: Local AI with no cloud dependency
- Team features: Shared sessions with roles and permissions
Try It Now
npx lecoder-mconnect
No installation needed. Just Node.js 20+ and cloudflared.
Start small: run a shell, scan the QR, type a command from your phone. Once you feel the power of mobile control, you'll never want to be tethered to your desk again.
---
Links
- GitHub: github.com/aryateja2106/lecoder-mconnect
- npm: npmjs.com/package/lecoder-mconnect
- Twitter: @r_aryateja
---
Built by LeSearch AI · Open Source · MIT License