Build guide

Claude Code Setup for Complete Beginners (No Code Required)

Install Claude Code from zero and run your first agent. No code required.

Claude Code
Mar 16, 2026
Optimized for

What you'll build

A working Claude Code setup on your own machine - installed, logged in, running inside a free IDE, and scoped to a single workspace folder. By the end you'll have your first agent ready to extend with skills.

What you'll need

Step 1: Get Claude Pro

Subscribe at claude.ai/upgrade. Pro is what authorizes Claude Code on your machine. The free tier won't work for this.

Step 2: Install Claude Code in your terminal

Open the Claude Code install page and copy the install command for your OS.

Paste the install command and hit enter. It downloads silently - no output is fine.

Step 3: Log in

In the terminal, type:

claude

It'll ask you to trust the folder - say yes. Then type login and pick the subscription option. It opens a browser, you authorize, you're in.

Step 4: Install Antigravity (your IDE)

Claude Code runs in the terminal, but you'll want a real interface to see files change in real time. Use Antigravity - Google's free IDE, very similar to VS Code. Download, install, open.

VS Code works too if you already have it. Either is fine.

Step 5: Create a scoped workspace folder

In Antigravity: File → Open Folder → create a new folder anywhere (call it whatever - axiom, claude-test, doesn't matter).

This folder is your agent's entire world. Claude Code can only read, write, and run commands inside it. Nothing outside this folder is reachable - which is why it's safe to give it broad permissions.

Step 6: Install the Claude Code extension + enable bypass permissions

Inside Antigravity:

  1. Go to Extensions → search "Claude Code for VS" → install
  2. Click the gear icon → toggle Allow dangerously skip permissions ON

Sounds aggressive. It's not - the agent is locked to this one folder. Without this toggle, Claude stops to ask permission on every small action and the productivity dies.

Step 7: Build your CLAUDE.md (the agent's brain)

CLAUDE.md is the only file Claude re-reads on every session. It's the agent's identity, rules, and context.

Fastest path to a good one:

  1. Open Claude Code Best Practices, copy the entire page
  2. In the Claude Code chat panel, paste it
  3. Then ramble - voice or typed - about what this agent is for. "My name is Austin, I want you to be my personal assistant, track my email each morning, brief me on finance and AI news..." Keep going for 2-3 minutes
  4. Hit enter. Toggle Plan Mode so it asks clarifying questions before generating

It writes your CLAUDE.md based on best practices + your context. Done.

Step 8: Add two universal skills

Skills are reusable automations the agent invokes by name. Two to install on day one:

Grab both from the free starter repo - copy/paste into your skills folder.

Step 9: Understand the model + token rules

Three models, picked automatically or via /model:

Each session has a 200k token context window. CLAUDE.md gets re-read on every message - keep it lean. When you hit the limit it auto-compacts and keeps going. Not a hard stop.

What to build next

Pick one process you do weekly or monthly. Email triage, expense tracking, content drafting, meeting summaries - whatever's repetitive. That's your first real skill. Use /skill-builder to make it.

Resources

Some links in this guide are affiliate links. They cost you nothing and support the free guides.

← Back to the library Good luck and happy building