AI Commander
TeamViewer for AI Agents, No VPN or SSH
Last verified:
What is AI Commander?
AI Commander is a remote shell for AI agents that lets Claude, ChatGPT, Codex, and any other AI assistant run real shell commands on machines you own. You install a small app on any computer—server, cloud VM, Raspberry Pi, Mac, or Windows PC—and get a stable machine code like AIC-7K3P-WX9M-RTBN. Then you share that code with your AI tool and ask in plain language what to do. The command runs on the target machine and the output streams back to you, with no exposed SSH, no open ports, and no VPN required.
Key features include three connection methods: MCP (simplest for Claude, Codex, Cursor, ChatGPT), API (plain web requests for scripts or scheduled jobs), and SKILL.md (for AI agents supporting skills). The agent keeps one outbound WebSocket open, working behind NAT, on headless boxes, and across fleets. Commands and results pass through encrypted and are never logged or saved. Fresh machine codes work for the first hour without an account; signing in enables ongoing access and fleet management with names like prod-db.
AI Commander is designed for developers managing cloud VMs, enterprise teams patching fleets, home users operating Raspberry Pis or NAS backups remotely, on-call engineers responding to incidents from their phone, and anyone running GPU/ML training boxes or debugging CI runners. It replaces VNC/port forwarding, bastion hosts, and screen-share calls. Use cases include restarting stuck services, freeing disk space, tailing logs, applying security updates across Linux boxes, checking SMART health on NAS, and updating IoT/edge devices behind cellular or NAT.
The tool works with Linux via a one-command install (cuommand install (curl -fsSL https://aicommander.dev/install | sudo bash), and offers desktop apps for Apple Silicon (M1–M4) and Intel Macs plus Windows x64 tray apps. It's built for doing the work rather than watching a screen, delivering text output you can read instead of a screen to watch. Security is built into the core: nothing exposed, codes never stored as-is, access that refreshes and expires, and results that are treated as plain text to show rather than new orders to follow
mljar
.
AI Commander pricing
Pricing model: Freemium
Fresh machine codes work for anyone for the first hour without an account. You do not need to sign in to try it. Sign in when you want ongoing access, keep machines in one place, and give them names like prod-db. The website does not list paid plans or pricing tiers—only that a free trial hour is available and sign-in enables continued access and fleet management mljar .
AI Commander pros
- No exposed SSH or open ports needed
- Works behind NAT and firewalls without setup
- Headless operation—no screen required
- Outbound-only WebSocket connection
- Commands and results never logged or saved
- Codes expire automatically and refresh on their own
- Works with Claude, ChatGPT, Codex, and more
- MCP connection with one command setup
- API option for plain web requests
- SKILL.md support for agent skills
- Fresh codes work for 1 hour without account
- Sign in enables fleet naming and management
- Same setup for Raspberry Pi, cloud VM, or fleet
- Linux one-command install
- Desktop apps for Apple Silicon and Intel Macs
- Windows x64 tray app included
- Replaces VNC and port forwarding
- No bastion host or VPN required
- Text output you can read, not screen to watch
- Machines stay separate from each other
- Codes are encrypted and never stored readable
- Cancel any key anytime
- Easy to try with safe default settings
- Hard to trick—logs treated as results not orders
AI Commander cons
- Only works on Linux, Mac, and Windows
- Requires installing an app on each machine
- AI gets real access to your machine—not sandboxed
- Codes must be kept private like passwords
- No inbound connections means some advanced setups limited
- Account needed for ongoing access beyond 1 hour
- MCP setup required for some AI tools
- Desktop apps only for Apple Silicon/Intel and Windows x64
- True fleet management requires signing in
Frequently asked questions about AI Commander
What is AI Commander?
AI Commander is a remote shell for AI agents. It lets Claude, ChatGPT, Codex, or any AI assistant run real shell commands on machines you own. You install a small app on a server, cloud VM, Raspberry Pi, Mac, or Windows PC, get a machine code, share it with your AI tool, and ask in plain words. The command runs and output streams back—with no exposed SSH, no open ports, and no VPN mljar .
How do I install AI Commander?
On Linux (cloud server, home server, or Raspberry Pi), paste one line: curl -fsSL https://aicommander.dev/install | sudo bash. On Mac or Windows, install the desktop app (Apple Silicon M1–M4 and Intel x64 for Mac; Windows x64 tray app). The app runs quietly in the background and gives you a stable machine code like AIC-7K3P-WX9M-RTBN mljar .
How do I connect my AI tool to AI Commander?
You have three options: MCP (simplest for Claude, Codex, Cursor, ChatGPT—run one command like claude mcp add --transport http aicommander https://aicommander.dev/mcp), API (send code and command to a web address for scripts or scheduled jobs), or SKILL.md (drop a ready-made Skill into any AI agent supporting skills). Most AI tools connect with one command, or you can just ask in plain words mljar .
Does AI Commander require open ports or SSH?
No. The app only reaches out via an outbound WebSocket—nothing on your machine waits for incoming connections. There's no open port to find or attack, and it works behind firewalls untouched. You don't need exposed SSH, bastion hosts, or VPNs aicommander +1 .
What machines does AI Commander support?
AI Commander works on Linux (servers, cloud VMs, Raspberry Pi), Mac (Apple Silicon M1–M4 and Intel x64 via desktop app), and Windows x64 (tray app). It's designed for headless boxes, NAT environments, and fleets across production servers aicommander +1 .
How long does a machine code work without an account?
A brand-new code works for anyone for the first hour without an account. After that, only approved accounts can use it. Sign in for ongoing access, fleet management, and to give machines names like prod-db aicommander +1 .
Is AI Commander secure?
Yes. Security is built into the core: nothing exposed (outbound-only), codes never stored as-is (encrypted), access that refreshes and expires, commands/results never logged or saved, machines stay separate, and results are treated as plain text to show—not new orders. Your code is the key—keep it private like a password. You can cancel any key anytime mljar .
What use cases does AI Commander support?
Home: restart stuck services, free disk, check Pi temperature, manage Docker/media server/Pi-hole, check NAS SMART health/backups. Developers: operate cloud VMs without SSH, drive GPU/ML training boxes, debug CI runners live. Enterprise: bulk patch fleets, install software across many machines, on-call incident response from phone, manage IoT/edge devices behind NAT/cellular aicommander .
How does AI Commander compare to SSH or TeamViewer?
AI Commander is built for AI and works with Claude/ChatGPT/Codex by design. It needs no open ports (SSH needs a port), works headless without a screen (TeamViewer needs a screen), works behind firewalls, and returns text you can read instead of a screen to watch. SSH requires extra setup behind firewalls; AI Commander works out of the box mljar .
Can I manage multiple machines with AI Commander?
Yes. Sign in to manage your fleet, keep machines in one place, and give them names like prod-db. You can apply security updates across Linux boxes, roll out agents/CLI/config changes to dozens of servers, and confirm everything landed from one conversation. Fleet patching and software installation work unattended without per-machine login aicommander +1 .