Overview
VibeCody can be deployed on any platform — cloud, desktop, or edge device — with a single command. This demo walks through the interactive setup wizard, always-on service installation, and multi-platform deployment. Think of it as your own self-hosted AI coding assistant, like myclaw.ai but free and running on your hardware.
Time to complete: 5–15 minutes depending on platform
Prerequisites
- A machine running macOS, Linux, Windows, or a Raspberry Pi 3/4/5
- (Optional) Docker installed for container-based deployment
- (Optional) Cloud CLI tools (aws, gcloud, az) for cloud deployment
Step-by-Step Walkthrough
Step 1: Run the Setup Wizard
The setup wizard auto-detects your platform, RAM, GPU, and recommends the optimal configuration.
vibecli --setup
Expected output:
┌─ VibeCody Setup Wizard ─────────────────────────────────┐
│ │
│ ✓ Platform: macOS (aarch64) │
│ ✓ Memory: 36.0 GB │
│ ✓ GPU: Apple Silicon (Metal) │
│ ✓ Hostname: macbook-pro │
│ ✓ Tier: max (recommended) │
│ │
└──────────────────────────────────────────────────────────┘
? Choose your AI provider:
▸ ollama Local models — free, private, no API key needed
claude Anthropic Claude — best for complex coding tasks
openai OpenAI GPT — widely used, fast
gemini Google Gemini — good free tier
grok xAI Grok — fast, generous rate limits
groq Groq — ultra-fast inference for open models
Enter choice (1-6, default 1): 1
✓ Recommended model for your hardware: codellama:13b
? Pull codellama:13b now? (this may take a few minutes) [Y/n] y
Running: ollama pull codellama:13b
? Enable always-on mode (run VibeCody as a background service)? [y/N] y
✓ Created ~/Library/LaunchAgents/com.vibecody.vibecli.plist
✓ Service loaded — VibeCody is running at http://localhost:7878
Checking VibeCody health... ✓ Healthy
┌─ Setup Complete ────────────────────────────────────────┐
│ │
│ ✓ Platform: macOS (aarch64) │
│ ✓ Provider: ollama (codellama:13b) │
│ ✓ Always-on: http://localhost:7878 │
│ │
│ Next steps: │
│ vibecli # Start chatting │
│ vibecli --agent "fix bugs" # Run an agent │
│ vibecli --review # Review code │
│ │
└─────────────────────────────────────────────────────────┘
Step 2: Manage the Background Service
Once installed, use the /service REPL command or --service CLI flag to manage the always-on daemon.
# Check status
vibecli --service status
# Stop the service
vibecli --service stop
# Start the service
vibecli --service start
Or from the REPL:
vibecli> /service status
● vibecody.service - VibeCody AI Coding Assistant
Active: active (running) since Fri 2026-04-04 10:30:00 PDT
vibecli> /service stop
✓ VibeCody service stopped
Step 3: Deploy to a Cloud Platform
For always-on team use, deploy to any cloud provider with one command.
# Clone the repo
git clone https://github.com/TuringWorks/vibecody.git
cd vibecody
# AWS (ECS Fargate)
./deploy/aws/setup.sh --tier pro
# Google Cloud (Cloud Run)
./deploy/gcp/setup.sh --tier pro
# Azure (Container Apps)
./deploy/azure/setup.sh --tier pro
# Oracle Cloud (FREE tier!)
./deploy/oracle-cloud/setup.sh --tier lite
# DigitalOcean
./deploy/digitalocean/setup.sh --tier lite
# Linode/Akamai
./deploy/linode-akamai/setup.sh --tier lite
Each script will:
- Validate cloud CLI authentication
- Provision infrastructure (container service, storage, networking)
- Deploy VibeCody with Ollama sidecar
- Print the access URL
Step 4: Deploy on a Raspberry Pi
VibeCody runs on Raspberry Pi 3, 4, and 5 — perfect for a privacy-first homelab AI assistant.
# On the Raspberry Pi:
curl -fsSL https://raw.githubusercontent.com/TuringWorks/vibecody/main/deploy/raspberry-pi/setup.sh | sh
The Pi setup script:
- Detects your Pi model and available RAM
- Installs the ARM64 binary
- Recommends the right model size (TinyLlama for Pi 3, Mistral 7B for Pi 5)
- Sets up a systemd service for always-on operation
- Optionally configures Cloudflare Tunnel or Tailscale for remote access
Step 5: Access from Anywhere
Once VibeCody is running as an always-on service, connect from anywhere:
Via direct HTTP:
curl http://your-server:7878/health
Via Tailscale (private network):
vibecli --serve --tailscale
# Access from any device on your Tailscale network:
curl https://vibecody.your-tailnet.ts.net/health
Via messaging gateway (Slack, Discord, Telegram):
vibecli gateway enable telegram
vibecli gateway enable discord
vibecli gateway enable slack
Now you can chat with VibeCody from your phone via Telegram, your team via Slack, or your community via Discord.
Step 6: Connect Multiple Platforms
Set up VibeCody on multiple devices with shared memory:
# On your Mac (main workstation)
vibecli --setup
# On your Raspberry Pi (always-on home server)
./deploy/raspberry-pi/setup.sh --always-on
# On AWS (team CI/CD integration)
./deploy/aws/setup.sh --tier pro
Deployment Matrix
| Platform | Setup Time | Cost | Always-On | GPU | Local Models | Remote Access |
|---|---|---|---|---|---|---|
| macOS | 2 min | Free | launchd | Metal | Up to 70B | Tailscale |
| Linux | 2 min | Free | systemd | CUDA/ROCm | Up to 70B | Tailscale |
| Windows | 3 min | Free | Service | CUDA | Up to 70B | Tailscale |
| AWS | 10 min | $15-60/mo | Auto | — | Via Ollama | Public URL |
| GCP | 10 min | $10-50/mo | Auto | — | Via Ollama | Public URL |
| Azure | 10 min | $15-55/mo | Auto | — | Via Ollama | Public URL |
| Oracle Cloud | 10 min | $0 | Auto | — | Via Ollama | Public URL |
| DigitalOcean | 5 min | $12-48/mo | Auto | — | Via Ollama | Public URL |
| Linode | 5 min | $12-48/mo | Auto | — | Via Ollama | Public URL |
| Raspberry Pi 5 | 5 min | Free | systemd | — | Up to 7B | Tunnel |
| Raspberry Pi 4 | 5 min | Free | systemd | — | Up to 3-7B | Tunnel |
| Raspberry Pi 3 | 10 min | Free | systemd | — | TinyLlama | Tunnel |
Demo Recording
{
"meta": {
"title": "Easy Setup & Deployment",
"description": "Deploy VibeCody anywhere with one command — cloud, desktop, or Raspberry Pi.",
"duration_seconds": 300,
"version": "1.0.0"
},
"steps": [
{
"id": 1,
"action": "shell",
"command": "vibecli --setup",
"description": "Run the interactive setup wizard",
"delay_ms": 15000
},
{
"id": 2,
"action": "Narrate",
"value": "The setup wizard detected macOS with Apple Silicon, 36 GB RAM, and Metal GPU. It recommended the max tier and codellama:13b model."
},
{
"id": 3,
"action": "shell",
"command": "vibecli --service status",
"description": "Check that the always-on service is running",
"delay_ms": 2000
},
{
"id": 4,
"action": "shell",
"command": "curl -s http://localhost:7878/health | jq .",
"description": "Verify the health endpoint responds",
"delay_ms": 2000
},
{
"id": 5,
"action": "Narrate",
"value": "VibeCody is running as an always-on service. Now let's deploy to a cloud platform for team access."
},
{
"id": 6,
"action": "shell",
"command": "cd deploy/digitalocean && ./setup.sh --tier lite --dry-run",
"description": "Preview a DigitalOcean deployment",
"delay_ms": 5000
},
{
"id": 7,
"action": "repl",
"commands": [
{ "input": "/service status", "delay_ms": 2000 }
],
"description": "Check service status from the REPL"
},
{
"id": 8,
"action": "repl",
"commands": [
{ "input": "/gateway enable telegram", "delay_ms": 3000 }
],
"description": "Enable Telegram gateway for mobile access"
}
]
}
Troubleshooting
| Problem | Solution |
|---|---|
| Setup wizard shows 0 GB RAM | Ensure VibeCLI has permission to read system info (sysctl on macOS, /proc/meminfo on Linux) |
| Ollama install fails | Install manually from https://ollama.com/download |
| Service won’t start | Check logs: journalctl --user -u vibecody (Linux) or cat ~/.vibecli/vibecli-stderr.log (macOS) |
| Cloud deploy fails auth | Run aws configure / gcloud auth login / az login first |
| Pi runs out of memory | Use a smaller model or add swap: sudo fallocate -l 2G /swapfile && sudo mkswap /swapfile && sudo swapon /swapfile |
| Health check fails | Verify port 7878 is not in use: lsof -i :7878 |
What’s Next
- Easy Setup Guide — Full platform comparison and one-command setup
- Use Cases — 80+ things to do with VibeCody
- Demo 55: Voice & Pairing — Voice control and pair programming
- Demo 14: Cloud Providers — Cloud IaC generation