Installation

Pi is a terminal-based agent — there’s no separate app to download. You’ll install it from the command line, and you’ll interact with it from the command line too. If you’ve never used a terminal before, that’s okay: Pi will walk you through everything, and you’ll pick up what you need as you go.

Quick system check

Install Pi

curl -fsSL https://pi.dev/install.sh | sh

npm install -g --ignore-scripts @earendil-works/pi-coding-agent

powershell -c "irm https://pi.dev/install.ps1 | iex"

npm install -g --ignore-scripts @earendil-works/pi-coding-agent

npm install -g --ignore-scripts @earendil-works/pi-coding-agent

You’ll need Node.js if you’re installing via npm/pnpm/bun.

Watch a quick walkthrough

Start Pi

cd /path/to/a/project
pi

Get connected to a model

Pi doesn’t ship with a free built-in model the way some tools do, so you’ll need to connect one before your first prompt works. Follow this decision tree:

1. Already have ChatGPT Plus or Pro? Run /login inside Pi and choose ChatGPT (Codex). This is officially sanctioned by OpenAI for third-party tools like Pi — safe to use.
2. No ChatGPT subscription? Sign up for a free OpenRouter account, grab an API key, and use one of the models tagged :free (for example google/gemma-4-31b-it:free). Zero cost, no local compute required.

A note on other subscriptions. Pi’s /login also lists Claude Pro/Max and GitHub Copilot as options. We don’t recommend either: Anthropic’s own terms explicitly prohibit routing Claude subscription credentials through third-party tools, and GitHub’s terms treat non-official Copilot clients as a violation too — both have real enforcement history (account bans/suspensions). Stick with ChatGPT Codex or an API key.

Send a test prompt

Hello, are you there?

If you get a response, you’re connected and ready to go.

Let's work on the "Installation" lesson from Pi School.