Setup — Claude Code¶
Ten minutes, three steps. If anything goes wrong, bring it to the break (10:15–10:40) — students will be on hand to help.
1. Install¶
Verify it worked:
2. Get an account¶
You need one of:
| Option | Cost | Good for |
|---|---|---|
| Claude Pro | $20/month | Most people. Generous Claude Code usage included. |
| Claude Max | $100–200/month | Heavy daily use, larger models, more parallel sessions. |
| API credits | Pay per token | Occasional use; you control spend exactly. |
For the seminar, Pro is plenty. Sign up at claude.ai — you can do this during the break.
3. Authenticate and run¶
In any terminal:
The first run opens a browser window to log in. When you see the prompt, you're in — Claude Code is a conversation that lives in your terminal and can read files, write files, and run commands in the directory you started it from.
Try:
Watch it run real commands (uname, which python, …) and report back.
That loop — model proposes, tool executes, model observes — is the entire
trick. Everything else in this tutorial is about making that loop work
harder for you.
Optional but recommended¶
- uv — fast Python package manager; the
agent will use it if present:
curl -LsSf https://astral.sh/uv/install.sh | sh - tmux — lets agents keep running when you close your laptop
(we cover this in Where Agents Run):
sudo apt install tmux/brew install tmux - A GitHub account with GitHub Education if you have a university email — free Copilot and more.
Permissions
Claude Code asks before running commands or editing files. During the seminar we'll loosen this (auto-accept mode) for speed. On your own machine, read what it asks the first few times — building a model of what the agent wants to do is part of learning the instrument.
Next: run your first "thing" →