Four tools, from “no install needed” to “AI coding agent on your own machine.” Pick one, start today.
AI for Dummies Module 4
You don’t need all four. Pick the one that fits where you are right now. The rest will still be here when you’re ready.
By the end of this lesson you’ll have at least one tool installed and working — ready for Lesson 05 (prompting).
Beginner path: Claude.ai first. Add the others only when you hit a task Chat can’t do on its own.
Cloud AI (Claude.ai, Claude Code) runs on Anthropic’s servers — your machine specs don’t matter. Local AI (LM Studio & Ollama) runs on your RAM, so this matters:
Step 1: Go to claude.ai in any browser — Chrome, Firefox, Safari, Edge.
Step 2: Sign up with your email or a Google account. Takes under a minute.
Step 3: Start typing. That’s the entire setup.
Free tier is enough to learn on. Desktop & mobile apps available — same login, same experience.
Recommendation: Start free. If you hit usage limits while doing real work, upgrade then — not before. The free tier handles learning perfectly.
The Discover tab — a built-in model browser:
1. Download from lmstudio.ai — Mac, Windows, or Linux. Installs like any normal app.
2. Open the Discover tab. Find a 3B–8B model. Click Download.
3. Open Chat, load the model, start talking — fully offline.
Bonus — local API server: LM Studio can serve a local API on port 1234 (OpenAI-compatible format). Apps that normally call a cloud API can point at your machine instead — zero cost, zero data leaving home. You don’t need this on day one; it’s there in Module 6.
The one-line rule: LM Studio is friendlier (clicks, no terminal). Ollama is leaner and better for scripting. Most beginners try LM Studio first, then discover Ollama when they want to automate something.
Install once, then run any open model with a single command. First run downloads the model (a few GB); after that it loads in seconds.
Type /bye to exit. On Linux use the one-liner above. On Mac/Windows download from ollama.com/download.
Memorize these four and you’ll cover 90% of what Ollama can do:
ollama run llama3.2 # download (first time) + start a chat
ollama list # see every model you've downloaded
ollama pull mistral # download a model without running it
ollama rm llama3.2 # delete a model to free up disk space
ollama serve # run Ollama as a background API serverPopular starter models:
llama3.2 — Meta’s fast 3B model
mistral — Mistral 7B, excellent balance
phi3 — Microsoft’s small but smart model
Local API: Ollama automatically exposes an OpenAI-compatible API on port 11434. Your own scripts can call it just like the cloud API — free, no data leaving home. See Module 6.
Claude Code is a coding agent. You run it inside a project folder and describe what you want in plain English. It reads your files, writes or edits code, and runs commands on your behalf.
What it can do:
What you need first:
Simplest. No extra software needed:
curl -fsSL https://claude.ai/install.sh | bashWorks on macOS and Linux out of the box.
If you already have Node.js 18 or newer:
npm install -g @anthropic-ai/claude-codeOn Windows: use inside WSL (Windows Subsystem for Linux).
After installing, cd into any project folder and type claude. It will open a browser window to authenticate on first run. That’s it — you’re running an AI coding agent.
Claude Code can edit real files and run real commands. It is not a chat window — it has actual access to your project on disk.
Think of it as a junior developer you’re mentoring: talented, fast, but you review the work before it goes live. Read what it proposes before approving. More on agent safety in Module 9.
| Tool | Runs where | Install? | Privacy | Best for |
|---|---|---|---|---|
| Claude.ai | Anthropic’s cloud | No | Data sent to Anthropic | Everyone, day one |
| LM Studio | Your machine (GUI) | App download | Fully offline | Private AI, no terminal |
| Ollama | Your machine (CLI) | Yes (one command) | Fully offline | Scripting, servers, devs |
| Claude Code | Your machine + cloud | Yes (curl or npm) | Code sent to Anthropic | Coding, multi-step tasks |
All four install commands in one place: see the Install Commands Gallery in the Theory page — copy buttons included.
Unsure about your situation? Use the Tool Picker Wizard in the Theory page — it asks 3 questions and gives you the exact install command.
Week 1 — Chat only. Use Claude.ai every day for real tasks. Learn what good output looks like. Practice asking — don’t install anything else yet.
Week 2 — Go local. Install LM Studio, download a 3B–8B model, have the same conversation you had on Claude.ai. Compare quality.
Week 3 — Touch the API. Run the Python example from Module 6. Process a small batch of something you actually care about.
Week 4 — Build one tiny thing. Pick a starter project from Module 8 and finish it.
You don’t need to rush. Each step works fine on its own. The later ones just unlock more. The single best thing you can do today: sign up for Claude.ai and use it for something real.
# Ollama (Linux/WSL)
curl -fsSL https://ollama.com/install.sh | sh
# Claude Code
curl -fsSL https://claude.ai/install.sh | bash
# Or via npm
npm install -g @anthropic-ai/claude-codeInstall one tool. Use it for a real task. Take a screenshot for the homework check.
Prompting — how to talk to AI well. The skill that improves your results more than any tool choice.