PiloTY MCP server

PiloTY is a helper that gives your AI its own live terminal window. Instead of running one command and forgetting everything, the AI can start something, walk away, and pick up where it left off. It is handy if you or your AI already run commands on a computer and want that to feel less like starting over every time.

What this MCP server does

You ask your AI to run something, like a build or a long test. The AI uses PiloTY to open a real terminal session and types the command in. PiloTY keeps that session alive, so the AI can check on it later, send more input, or stop it. When you ask again, the AI reuses the same session, so the folder it is in, the settings, and anything already running are still there. You get the output back in your chat.

Click to zoom

What you can do with it

• Start a long job like a build or test run and check on it later
• Watch a log file that keeps updating, such as tail -f
• Keep a Python or debugger session open while your AI reads code
• Type passwords when a command asks for them, like sudo or SSH
• Stay logged in to a remote server across several steps
• Look at what is on the screen, not just the text, for full-screen programs
• Stop a running command with an interrupt signal

Try asking your AI

• “Start the test suite in my project folder and tell me when it finishes”
• “Open a terminal in /home/me/app and run the build, then show me the last lines”
• “Log in to my server over SSH and check how much disk space is left”
• “Keep a Python session open and try a few lines of code for me”

What it gives back to you

You get the text the terminal printed, plus a short note on how the call ended, like success or ran out of time. It also tells you what state the terminal seems to be in, for example running, waiting for a password, or sitting at a prompt. For full-screen programs you can ask for a snapshot, which shows what a person would see on the screen. Everything is shown to you in the chat.

Before you start

Who it's for

Developers and technical folks who already use an AI assistant to run commands and want those commands to keep their state between steps.