This is a beginner’s guide to getting started with Claude Code, the AI tool I
use most — written by Bryan Cutsinger, an economist at Florida Atlantic
University (and interim director of the AIER Sound Money Project), for other
economists. The examples lean on academic life: cleaning a dataset, tightening a
working paper, drafting a problem set. No prior experience needed — we start from
zero and build up. It’s free, self-paced, and updated as the tools change: no
sign-up, no paywall, no consultation gate. (For Anthropic’s own reference, see the
official quickstart.)
The short on-ramp: decide whether Claude Code is for you, then do the one-time setup.
What Claude Code is (and isn't)A three-minute orientation: what it does, how it differs from chatting with Claude in a browser, and the one rule that governs everything — you stay the reviewer.
Set up Claude CodeA one-time, beginner-friendly walkthrough: install the tool from the Terminal, sign in, fix the common snags, then add VS Code and the Claude Code extension as your home base.
The core loop and how to steer it — no new tools needed. A reader who stops at the end
of this group is productive.
Your first sessionRun Claude Code end to end on a safe practice folder — ask it a question, approve your first change, and learn to manage and hand off a conversation.
Customizing Claude with a CLAUDE.md fileThe single highest-leverage habit: a file that tells Claude who you are, how you work, and your house style — so you stop re-explaining yourself every session.
Controlling ClaudeHow to steer a session on real work: the explore, plan, implement, commit loop, plan mode, the permission modes, and the habit of staying the reviewer.
Models, cost, and usage limitsChoosing a model and managing what it costs: the model tiers, effort and fast mode, the two billing models, session and weekly limits, and habits that keep usage down.
Getting material into ClaudeThree ways to feed Claude well: dictating long prompts, converting documents to Markdown so Claude reads them cleanly, and pasting figures, screenshots, and PDFs.
Next-stage reading for when you point Claude at a real paper, dataset, or course — the
two “before you touch real data” chapters first.
Working safely on real projectsHow folder scope limits what Claude sees, what to keep out, working on a copy, and using rewind to undo a change you didn't want.
Keeping your data privateWhat happens to data you give Claude, why your plan choice matters, and how to work on confidential or IRB-governed data without exposing it.
Working in VS CodeSettle into VS Code for real work: juggle multiple integrated terminals, and add the few extensions an economist needs.
Version control: Git & GitHubWhat Git and GitHub are in plain terms — an undo history for a whole project, plus a cloud home for it — and the handful of moves you'll actually use.
Writing papers & slidesClaude Code inside a real academic writing workflow: a LaTeX paper kept in sync between Overleaf and GitHub, Beamer slides, and a verify-every-citation discipline.
Come-back-later material, ordered by rising power. API keys come first because the
chapters after them lean on credentials.
API keys and secretsWhat an API key is, when you actually need one, and how to handle keys safely — never in a file, a commit, or printed output.
Connecting Claude to your appsApp connectors and the browser extension on the Claude side, and MCP servers that plug data sources like Zotero or a database into Claude Code.
The Claude app and the cloudThe Claude desktop and web apps and cloud sessions — what they add over the terminal, and when your data should and shouldn't leave your Mac.
Sub-agentsSeparate Claude workers you hand a focused job — and how to use them to keep your main session clean.
Rules, hooks, and skillsThe customization layer beyond CLAUDE.md: hooks that run on their own to enforce your guardrails, and skills that package a repeatable job into a single command.