AI CLI Tools for Non-Engineers: A Current Getting-Started Guide

AI CLI tools are assistants and coding agents that work from a terminal. You describe a task and, depending on the tool and its permission settings, it can inspect a project, propose or make edits, run commands, and report the result. They do not remove the need to define scope, approve risky actions, and check the work.

This guide is for people who have never opened a terminal or installed a developer tool. You will learn enough to start in a disposable project folder, compare three current tools, and choose a path that fits your account and budget. The examples are starting points, not a promise that every project will work without setup, review, or paid usage.

What Is a Terminal (and Why Should You Care)?

You already know how to talk to your computer. You click things, drag things, and tap buttons. That is the graphical way: icons, windows, and menus. A terminal is another interface. You type a command and get text back, which makes it useful for repeatable work and for tools that need access to a local project.

A normal browser chat usually returns text in the browser. A CLI tool is designed to work in the directory where you start it, so it can use local files and commands that you allow. That extra access is useful, but it also means you should start in the right folder, read permission prompts, and review changes before sharing or deploying them.

The Four Commands That Do Everything

You do not need to memorize a hundred commands. Start with four.

ls -- Look around. This shows the files and folders in your current location. It is like opening Finder on Mac or File Explorer on Windows and viewing a folder.

ls

cd -- Move to a folder. cd stands for "change directory." It moves you into a folder.

cd Documents

cd .. moves up one level. The two dots mean the parent folder.

mkdir -- Make a new folder. This creates a folder without changing anything else on your computer.

mkdir my-project

You can then run cd my-project to enter it.

pwd -- Where am I? This prints the full path to your current location.

pwd

You might see /Users/yourname/Documents/my-project. Check this before asking an agent to edit files.

That is enough to navigate a project by typing. The agent can help with many other commands, but you should still understand what directory it is using.

How to Open Your Terminal

• Mac: Press Cmd + Space, type "Terminal", and press Enter.
• Windows: Press the Windows key, type "PowerShell", and press Enter.
• Linux: Open the terminal application provided by your distribution.

A terminal will run the commands you approve. That is why an unfamiliar command deserves a pause, especially if it deletes files, changes system settings, downloads software, or sends data to a service.

Install Your First AI CLI Tool

This guide uses Claude Code for the first conversation. It is made by Anthropic and is designed to inspect and edit a project from the terminal. The current Claude Code setup documentation recommends a native installation. The native binary does not require a Node.js runtime.

Step 1: Install Claude Code with the native installer

The current Claude Code setup documentation recommends the native installer. The installed binary does not require a Node.js runtime, so Node is not a prerequisite for this path. Choose the command for your operating system, then open a new terminal window and verify the installation:

macOS or Linux (run in Bash or Zsh):

curl -fsSL https://claude.ai/install.sh | bash

Windows PowerShell:

irm https://claude.ai/install.ps1 | iex

Windows Command Prompt:

curl -fsSL https://claude.ai/install.cmd -o install.cmd && install.cmd && del install.cmd

On native Windows, installation does not require Administrator permission. If you use WSL, open the WSL terminal and use the macOS/Linux command there rather than running it from PowerShell or Command Prompt. After installation, open a new terminal and run:

claude --version

If the command is not found, follow the setup page's PATH troubleshooting instead of installing Node.js as a guess.

Alternative: npm installation

If the native installation worked, skip this section and sign in below. npm is an alternative installation route. If you choose it, the current official documentation requires Node.js 22 or later for the package installation. Check your version first:

node --version

Then install the package:

npm install -g @anthropic-ai/claude-code

The npm package installs the same native binary, which does not invoke Node.js at runtime. The -g flag installs the package globally for your user or configured npm prefix. Verify the command is available:

claude --version

Do not run sudo npm install -g; the official setup documentation warns that it can create permission and security problems.

Step 2: Sign in and understand the account boundary

Run Claude Code in the project folder:

claude

The current setup page says Claude Code access requires a Pro, Max, Team, Enterprise, or Console account, or a supported third-party provider such as Amazon Bedrock, Google Cloud's Agent Platform, or Microsoft Foundry. The free claude.ai plan does not include Claude Code access. Follow the browser prompt for the account path you use.

A Claude subscription and Anthropic Console/API usage are separate billing paths. Console usage uses API credits and standard API pricing; a subscription does not silently turn API calls into free calls. Check Anthropic's current pricing, Claude Code setup and authentication, and API credit guidance before choosing a plan. Prices, limits, and available features can vary by region and change over time.

Prefer Google's CLI?

For individual users, start with Antigravity CLI, not the old Gemini CLI free-account tutorial. Google's transition notice took effect on June 18, 2026 for free Gemini Code Assist for individuals and Google AI Pro/Ultra users.

Open the official quickstart, select the Antigravity CLI instructions for your operating system, and follow its installation and sign-in steps. After installation, verify the binary and launch it from your project folder:

agy --version
agy

Antigravity's current plans set its own usage limits. Do not carry over Gemini CLI's old 1,000/day or 250/day figures. Existing Gemini CLI users can follow the migration guide. Gemini CLI itself remains an option for supported Code Assist Standard/Enterprise and paid API routes, not the consumer free-account path described in older tutorials. Checked on October 27, 2026.

The rest of this guide uses Claude Code in its examples. Other clients use different commands and quotas, but the basic loop is similar: work in a known folder, describe the task, review permissions and changes, then verify the result.

What about Codex CLI?

OpenAI's current Codex CLI guide documents installation options for macOS/Linux, Windows, npm, and Homebrew. After installation, open a project directory, run codex, and choose Sign in with ChatGPT or another available authentication method. Use the official guide for the installation option that matches your operating system.

According to OpenAI's current ChatGPT plan guidance, Codex is included across ChatGPT plans including Free and Go, but usage limits vary by plan. Codex Cloud has a different eligibility boundary. API-key usage is a separate OpenAI API billing path, so a ChatGPT subscription should not be described as a general API credit balance.

Your First Conversation with an Agent

Start in a new folder so you know which files the agent can see:

mkdir my-first-project
cd my-first-project

Start Claude Code:

claude

You will see a prompt. Ask a small question first:

What can you do in this project, and what would you need permission to change?

The response should describe its available tools and boundaries. Read the answer instead of assuming that every tool, shell command, or external connection is enabled.

Now ask it to make a small static page:

Create a single index.html webpage about cats. Include a title and three fun facts. Use a modern, accessible layout and explain which files you plan to create before editing them.

Review the proposed files and commands. If the agent asks for approval, approve only what you understand. When it finishes, inspect the files and open the page in a browser:

• Mac: open index.html
• Windows: start index.html

A local HTML page is a useful first exercise because it has a small surface area. It is not proof that a larger application is secure, correct, or ready to publish.

Four Projects You Can Build Today

Each project below has a dedicated article with step-by-step guidance. Treat those articles as examples: dependencies, hosting terms, permissions, and output quality vary by project and provider.

From One Sentence to a Running App

Describe a small app idea such as a habit tracker, budget tool, or reading list. Ask the agent to propose the stack, create the project, explain the dependencies, and show how to run it. Review each generated file and command before you continue.

Read the full guide: From One Sentence to a Running App

Build and Deploy a Landing Page

You can describe a layout, color direction, and content, then ask an agent to produce HTML and CSS. Hosting providers may offer free tiers, but domains, bandwidth, build minutes, and other services can cost money. Check the provider's current terms before promising a free deployment.

Read the full guide: Build and Deploy a Landing Page

Turn Your Expertise into a Tool

Describe a rule set or workflow you understand well, such as a meal-planning calculator, a wine-pairing guide, or a language-learning exercise. Ask the agent to turn the rules into a small interface, then test edge cases against your own domain knowledge.

Read the full guide: Turn Your Expertise into a Tool

Organize Years of Scattered Notes

Export a copy of your notes into a working folder and start with a plan for classification. Keep the originals untouched. Ask the agent to propose a schema and a sample transformation before it rewrites a large set of files. Personal, confidential, or regulated information deserves an additional privacy review.

Read the full guide: Organize Years of Scattered Notes

Choosing the Right Tool

The options below have different providers, installation paths, sign-in methods, and billing boundaries. Start with the account you can use, the installation method for your computer, and the amount you are willing to spend.

For a broader, frequently changing survey of AI CLI tools, the AI CLI Tools Guide covers more products. Recheck each provider's official docs before installing or budgeting.

Common Fears (and Better Defaults)

"Will I break my computer?"

An agent started in a project folder can still run commands with effects outside that folder if you approve them. Start with a disposable directory, keep backups, read the command and file list, and avoid granting broad permissions until you know the workflow. A project folder is an organizational boundary, not a guarantee that the rest of your computer is unreachable.

"Do I need to learn programming?"

You can start without prior programming experience. You will get better results if you learn enough to identify the files, dependencies, data handling, and checks involved. You do not need to understand every line before asking for help, but you should be able to review what will run and what will be published.

"Is this expensive?"

There is no honest universal cost for a beginner project. Claude Code can use a subscription or Console/API credits; Google's consumer route follows Antigravity's current plan limits, while supported Gemini CLI enterprise/API routes have their own billing; Codex can use ChatGPT plan capacity or separate API credentials. Hosting, domains, and packages can add costs. Check the linked official pricing and quota pages before you start.

"What if the AI makes mistakes?"

It will sometimes produce incomplete, unsafe, or incorrect work. Ask it to explain the change, inspect the diff, run relevant tests, and compare important results with a source you trust. For a visual page, check it in the browser. For a data transformation, keep the original and compare a sample before processing everything.

"I tried coding before and hated it."

You can begin with plain-language requests instead of memorizing syntax. The terminal is still a technical environment, so expect to learn a few commands, read error messages, and make decisions about permissions and data. The goal is not to pretend those decisions disappear; it is to get help with the repetitive parts while you keep control of the outcome.

What's Next: From Beginner to Power User

Once the terminal feels familiar, you can add more structure. None of these steps is required for a first page.

Project instructions. Claude Code can read a CLAUDE.md file in a project to apply documented conventions. Keep instructions specific, reviewable, and limited to that project. The Claude Code documentation explains the current loading and configuration behavior.

MCP servers. Model Context Protocol connects an agent to external tools and data. Treat every connector as another permission and data-sharing boundary. Read its documentation, credentials policy, and allowed actions before enabling it.

Multi-agent workflows. Multiple agents can work on separate tasks, but parallel work adds coordination, permission, and review overhead. Start with one small task, define how outputs will be checked, and measure whether parallel work helps your project instead of assuming a productivity gain.

For a deeper discussion of multi-agent workflows, context engineering, and advanced AI CLI strategies, the AI Agent Workflow Guide covers the next level.

Try a Handoff in Termdock

You do not need many agents for your first try. Open a small project you are authorized to use in Termdock and start an installed CLI in a terminal. Ask it for one small change and a list of edits and test results. Afterward, open another terminal and use another installed CLI or a fresh session to check the change against the requirement; repeat the requirement and file locations because sessions do not share the conversation. Inspect the result yourself. For an agent-driven handoff, see the Termdock walkthrough. Each tool's sign-in and model costs are separate from Termdock.

A Practical Next Step

AI CLI tools can shorten the path from an idea to a local prototype, but they do not replace planning, permissions, source checking, testing, or human responsibility. Start with a small folder, make one change you can inspect, and only then add dependencies, external services, or deployment.

The four projects in this guide are starting points: a small app, a landing page, a domain-specific tool, or a searchable notes structure. Choose one with data you are allowed to use, keep a copy of the original, and let the agent help with the work you can verify.

Open the terminal. Type one clear request. Read what it plans to do.