How to Get Started with Claude Code: Installation, Pricing, and Token Saving

Command-line AI tools often feel intimidating if you are used to chatting with Claude in a browser window. Anthropic designed Claude Code to bridge that gap by bringing an agentic coding assistant directly into your terminal. Instead of copying and pasting code snippets back and forth, the tool can read your repository, execute tests, edit files, and answer questions about your project directly where your work lives.

Here is how Claude Code works, what plan you need to run it, and how to set it up without burning through your subscription or API credits.

Who Can Access Claude Code

Anthropic makes Claude Code available across paid subscription tiers and API accounts. You can run it if you have an active Claude Pro subscription ($20 per month, or $17 per month billed annually), a Claude Team plan ($25 per seat monthly), or a Claude Max plan ($100 to $200 per month). The standard Free tier does not include access to the CLI.

If you prefer paying strictly for what you use, you can connect an Anthropic API key instead of a subscription. With an API key, you pay per token based on the underlying model you call, such as Claude Sonnet 5.5 at $2.00 per million input tokens and $10.00 per million output tokens, or Claude Haiku 5.5 at $0.10 input and $0.50 output per million tokens for standard prompts. For most everyday tasks, running Claude Code on an existing Claude Pro or Team subscription provides predictable monthly budgeting without per-token surprise bills.

Installing the CLI the Native Way

Anthropic officially recommends using the Native Install script rather than package managers like npm. This keeps the installation lightweight and ensures your CLI stays on the current update path.

On macOS, Linux, or Windows Subsystem for Linux (WSL), open your terminal and run the official installation command:

If you are on Windows using PowerShell, you need Git for Windows installed so Claude Code can access required Bash-compatible utility commands. Once Git is set up, run this installer command in PowerShell:

Verifying and Logging In

Once the script completes, restart your terminal session or open a new tab so your path updates. You can confirm the installation succeeded by checking the version:

Anthropic also includes a built-in diagnostic command that verifies your shell permissions, network connectivity, and dependencies. Running this early saves troubleshooting time:

When you launch claude for the first time inside your project directory, the tool opens an interactive sign-in flow. It directs your browser to authenticate with your Anthropic subscription or prompts you to supply your API key. Once linked, the agent is ready to inspect your code.

Saving Tokens with a CLAUDE.md File

Because agentic CLI assistants read local files and analyze directory structures, they can consume context tokens rapidly if left undirected. Anthropic built an automatic project configuration standard to keep token usage lean.

Navigate to your project root and run the initialization command:

This command scans your repository and creates a lightweight CLAUDE.md markdown file in your root folder. This file serves as persistent memory for the CLI, capturing essential project details such as your package manager, testing commands, code formatting rules, and folder structure.

Claude Code automatically ingests CLAUDE.md every time it starts. Because the assistant immediately understands how your project builds and runs without having to re-scan every folder or ask repetitive clarifying questions, you save context tokens on every single query. It is a quick five-minute step that keeps your monthly rolling usage healthy.