What Is Cmd+Ctrl?

February 22, 2026 · Michael Richman

Does this sound familiar? You get up from your computer thinking your agent is working away. You come back an hour later ready to see what it has accomplished – only to realize it stopped after five minutes to ask you a question. It's been sitting there blocked, waiting for you the entire time.

That is the problem Cmd+Ctrl (pronounced "command and control") is designed to address.

The productivity paradox

AI coding agents promise to accelerate our work. Claude Code, Cursor, Codex, GitHub Copilot, Gemini CLI – they read files, write code, run tests, refactor modules. They work autonomously for minutes or hours at a time.

But here's the paradox: the more agents you run, the more time you spend babysitting them. Three to five simultaneous sessions, each one taking minutes or hours, each one needing input at unpredictable times. You can't get up from your damn computer.

You go to make coffee. You hear the notification sound. Gotta go check what it needs. You go for a walk – need to get outside, get some fresh air, think. You have ideas about what to do next. And you have to wait till you get back to act on them.

Even if we want to believe these agents are fully autonomous, the truth is that one-shot perfection is still a holy grail. Agents get stuck. They go off the rails. They need course corrections. They ask questions.

The Flow problem

There's a deeper issue here. We used to think of Flow as complete immersion in a single task – you and your IDE, locked in, solving one problem.

The new Flow – the one we're trying to achieve with multiple autonomous agents – is different. It's orchestration. It's choreography. Three, four, five agents working in parallel. You move between them, checking in, unblocking one, reviewing another's work, steering a third in a new direction.

Except we can't achieve it. We're switching between terminal windows. Agent one needs input. What was agent two doing? Did agent three finish or is it still running?

The agents are tightly coupled to their execution contexts. You run an agent in a terminal, you interact with it in that terminal. You run it in an IDE, you interact with it in that IDE. There's no way to step back and see the whole picture – let alone interact with it from somewhere else.

Where Cmd+Ctrl comes in

AI coding agents are asynchronous actors with lifecycle states – working, paused, blocked, done. They're long-running processes, not interactive sessions. And they need infrastructure that treats them as such.

A command and control system for AI agents needs three things:

1. Awareness – see what your agents are doing in real time, from any device. Which ones are working? Which ones are blocked? Which one have you forgotten about?

2. Control – send instructions from wherever you are. Reply to a question from your phone. Start a new task from your watch. Redirect an agent from your tablet.

3. Decoupling – agent execution happens in its native environment. Human interaction happens through a unified control plane. The two are independent.

How Cmd+Ctrl works

Cmd+Ctrl is a remote orchestration system for AI coding agents. A lightweight daemon runs alongside each agent on your dev machine, monitoring its lifecycle state and connecting to a server over WebSocket.

Your Phone / Browser / Watch
        ↕
   Cmd+Ctrl Server
        ↕
   Daemon (on your dev machine)
        ↕
   AI Agent (Claude Code, Cursor, etc.)

The daemon bridges the gap. It translates between the agent's native interface and Cmd+Ctrl's protocol, streaming tool usage, file reads, and responses back to you in real time. When you send a message from your phone, it routes through the server, to the daemon, to the agent.

The critical piece: this is bidirectional. The agent can signal to you, and you can respond back – all without being at the machine where the agent is running.

You get push notifications when an agent finishes work or asks a question. You can check in on a running agent mid-task to see if it's making progress or going off track. You can start new sessions, review progress, and reply – all from your phone while you're away from your desk.

And this works across agent frameworks. Same control plane for Claude Code, Cursor, Codex, GitHub Copilot, Gemini CLI – whatever you're using. No more hunting through terminal tabs trying to remember which agent is in which state.

After a one-line install, every session is visible and accessible from app.cmd-ctrl.ai.

Roam free

The goal isn't to be more chained to work. It's to be less chained to your desk. Take the walk. Get the coffee. Have the ideas. And when an agent needs you, it can let you know – wherever you are.

❯Get started

curl -fsSL https://docs.cmd-ctrl.ai/install.sh | bash

See the installation docs for full setup details.