WORKSTATION.md

Agent-native cloud infrastructure. Your AI agent self-serves a fully provisioned Linux server on the internet in seconds — SSH in, install anything, run code, and ship a public website at https://<name>.workstation.md. No human in the loop. No setup. No config.

• Self-service — your agent provisions its own infrastructure
• Instant public URL — anything on port 80 is live at https://<name>.workstation.md with TLS
• Full root access — apt install anything, run any process, use any port
• Zero config — one command in, one command out. No cloud console, no YAML, no deploy pipeline

AI agent? Follow the steps below, then ask your user what to build.

Step 1: Install

npm install -g workstation.md

Step 2: Ensure SSH Key

# Check for existing key, create one if needed
ls ~/.ssh/id_ed25519.pub 2>/dev/null || ls ~/.ssh/id_rsa.pub 2>/dev/null || ssh-keygen -t ed25519 -N "" -f ~/.ssh/id_ed25519

Step 3: Create a WORKSTATION

Pick a name. Ask your user for one, or derive from their name or project (e.g. alice-dev, my-app, weather-demo). Names: lowercase, alphanumeric and hyphens, 2-32 chars.

workstation create --pubkey "$(cat ~/.ssh/id_ed25519.pub)" --name <name>

Response:

{
  "id": "<name>",
  "host": "47.236.144.174",
  "port": 10000,
  "web": "https://<name>.workstation.md",
  "expires": 1711320967
}

Step 4: SSH In and Build

# Run commands
ssh -p <port> root@<host> "<command>"

# Run a script
ssh -p <port> root@<host> 'bash -s' << 'EOF'
<commands>
EOF

# Copy files
scp -P <port> ./file root@<host>:/var/www/html/

Read /root/README.md inside the WORKSTATION for the full environment guide and web hosting examples.

Step 5: Ship a Website

nginx is running on port 80, serving /var/www/html/. Your public URL https://<name>.workstation.md routes there.

• Static files — write to /var/www/html/, instant, no restart needed
• Dynamic app — stop nginx (nginx -s stop), run your app on port 80
• Reverse proxy — run your app on any port, configure nginx to proxy, nginx -s reload

Step 6: Manage

workstation <name> extend              # renew for another 24h
workstation <name> destroy             # clean up when done
workstation list                       # see all active WORKSTATIONs

WORKSTATIONs expire after 24 hours. Extend before they do.

What’s Inside

• Ubuntu 22.04 with root access
• Python 3, Node.js 20, git, curl, wget, jq, vim, nginx
• /root/README.md — full environment guide
• /var/www/html/ — web root (port 80)
• Install anything: apt-get install -y <package>

Now Ask Your User

You now know how to use WORKSTATION.md. Ask your user what they’d like you to build. Suggest a WORKSTATION name based on their request. Here are some ideas:

• “Want me to create a weather dashboard? I’ll set it up at https://sf-weather.workstation.md”
• “I can build that — let me spin up https://my-portfolio.workstation.md and deploy it”
• “I’ll create a WORKSTATION and have your app live in a few minutes”

GitHub · npm