---
title: Claude Code Explained: What It Is, Cost and Alternatives
url: https://www.masaischool.com/resources/claude-code-explained
description: What Claude Code is, where it runs, what it costs (not on the free claude.ai plan, 11 Nov 2026), a first-hour walkthrough, safe use and alternatives.
updated: 2026-11-17
publisher: Masai (https://www.masaischool.com)
---

# Claude Code Explained: What It Is, What It Costs, a First-Hour Walkthrough and Free Alternatives

> **Short answer**
>
> Claude Code is Anthropic's AI coding agent. It reads your project, edits files, runs commands and tests, and works from a terminal, VS Code or JetBrains, a desktop app or the web. It needs a paid Claude plan or Console API credits: as of 11 November 2026 the free claude.ai plan does not include it.

Searches for Claude Code in India are more than four times last year's level on Google Trends, and many answers are marketing or out of date. This page is a neutral guide: what Claude Code is, what you need to start, what is free and what is not, how to use it safely, and what else you can try.

The details come from Anthropic's documentation and pricing page, read on 11 November 2026. I did not run Claude Code to write this page, so check version numbers and menus on your own screen. Where a tool's free status is unclear, the table says so.

1. **Ask** - describe one small task
2. **Explore** - Claude reads your files
3. **Act** - it edits and runs, with your approval
4. **Check** - run tests, read git diff
5. **Commit** - you decide what to keep
A safe loop for using a coding agent.

## What Claude Code is, and where you can use it

Claude Code is an agentic coding tool from Anthropic. Its documentation says it reads your codebase, edits files, runs commands and connects to development tools. 'Agentic' means it works in a loop: it decides an action, runs it, reads the result and carries on until the task is done or it needs you. [What is an AI agent](https://www.masaischool.com/resources/what-is-an-ai-agent) explains the loop. A chat window answers. An agent acts on your files, which is why the safety section below matters.

| Where | What you get | Notes from the docs |
|---|---|---|
| Terminal (command-line interface, CLI) | The full-featured version. Type claude in a project folder | One-line installer for macOS, Linux, Windows and WSL (Windows Subsystem for Linux) |
| VS Code | An extension with inline diffs, plan review and history | It can also be installed in Cursor |
| JetBrains | A plugin for IntelliJ IDEA, PyCharm, WebStorm and other JetBrains development tools | The plugin needs the CLI installed separately |
| Desktop app | A standalone app to review diffs visually and run several sessions | A paid subscription is required. The Linux version is in beta for Ubuntu and Debian |
| Web | claude.ai/code in a browser, and the Claude app on iOS and Android | No local setup. Useful for repositories you do not have on your machine |

Every surface uses the same engine, so a project's CLAUDE.md instructions, settings and MCP servers work in all of them. [MCP](https://www.masaischool.com/resources/what-is-mcp-explained) is how Claude Code connects to outside tools such as GitHub or a database. The docs list building features from a plain-language description, fixing bugs from an error message, writing tests and making git commits and pull requests. Masai's overview is [Claude Code explained: what it is and how to use it](https://www.masaischool.com/blog/claude-code-explained-what-it-is-how-to-use-it/).

## What you need and what it costs, as of 11 November 2026

Claude Code needs an account. The setup docs say it requires a Pro, Max, Ultra, Team, Enterprise or Console account and that the free claude.ai plan does not include Claude Code access. The Claude pricing page agrees: Claude Code is not in Free and is in Pro and Max.

| Route | What you pay | Notes |
|---|---|---|
| Free claude.ai plan | $0 | No Claude Code access |
| Pro | $17 a month billed annually ($200 up front), or $20 a month billed monthly | Includes Claude Code. Prices are in US dollars, the page showed no INR price, and it lists no individual student discount (its Education plan is bought by universities) |
| Max | $100 a month | Listed as 5x the usage of Pro, with Claude Code included |
| Console (API) | Pre-paid credits, charged by tokens used | A 'Claude Code' workspace is created in the Console on first login. Claude Sonnet 5.5 lists at $2 per million input and $10 per million output tokens |
| Cloud providers | Your cloud bill | Amazon Bedrock, Google Cloud's Agent Platform or Microsoft Foundry |

How much does heavy use cost on the API? Anthropic's cost guide says that across enterprise deployments the average is about $13 per developer per active day and $150 to $250 per developer per month, with costs below $30 a day for 90% of users. That is company usage, not a student's, and yours will differ. Watch /usage as you work: it shows token use for API users and plan usage bars for subscribers.

So Claude Code itself has no free tier. Two low-cost routes exist. Anthropic's Claude Academy publishes free self-paced courses, including Claude Code 101. And Ollama documents a way to point the Claude Code interface at open-weights models on your own machine: set ANTHROPIC_BASE_URL=http://localhost:11434 and pick a model that supports tools and has a context window of at least 64K tokens. That is Ollama's setup rather than Anthropic's, and local models are usually less capable than Claude models.

## Install it and do a first-hour walkthrough

The setup docs list these requirements: macOS 13 or later, Windows 10 version 1809 or later, Ubuntu 20.04 or later, Debian 10 or later or Alpine Linux 3.19 or later, at least 4 GB of RAM, an internet connection and a country on Anthropic's supported list. The installer below is the one Anthropic documents. Read what any script does before you pipe it into a shell. Homebrew (brew install --cask claude-code) and WinGet (winget install Anthropic.ClaudeCode) are alternatives.# macOS, Linux or WSL curl -fsSL https://claude.ai/install.sh | bash # Windows PowerShell irm https://claude.ai/install.ps1 | iex # then, in a new terminal claude --version

Now make a practice project: a new folder, git init, and this file saved as grades.py. It runs and prints B, so you can confirm it works before any AI touches it.def average(marks): return sum(marks) / len(marks) def grade(avg): if avg >= 90: return "A" if avg >= 75: return "B" return "C" print(grade(average([80, 70, 78])))

Output: B. The bug is that average([]) raises a ZeroDivisionError. That is your first task.

1. Commit the starting point: git add . then git commit -m start. Now every change Claude makes shows up in git diff.
2. Open a terminal in the folder, run claude and log in when the browser prompt appears.
3. Ask a question first: 'what does this project do?' Claude reads the files and answers. Nothing changes.
4. Give a small task with a clear finish: 'average([]) crashes with ZeroDivisionError. Reproduce it, make it return 0, and add a test. Show me what you changed.' Anthropic's best-practices page says a prompt that names the symptom and says what 'fixed' looks like usually saves a round of corrections.
5. Read each permission prompt. Claude asks before editing files or running commands unless your mode lets it proceed. Read the command before you approve it.
6. Review the diff yourself with git diff, run the tests yourself, and only then commit. Ask Claude to explain any change you do not follow.
7. Try /init to generate a CLAUDE.md with notes about your project, /clear before an unrelated task, Esc to interrupt when it drifts, and /rewind to restore earlier file checkpoints.

When the practice task feels routine, follow [build an app with an AI coding agent](https://www.masaischool.com/resources/build/build-an-app-with-an-ai-coding-agent) for a larger project, and [write an AGENTS.md file](https://www.masaischool.com/resources/build/write-an-agents-md-file) to give any coding agent standing instructions. Claude Code can read an AGENTS.md in place of a CLAUDE.md.

## Safe-use rules

- **Use Git and commit before each session.**/rewind undoes changes to files Claude edited with its file tools. A file changed only by shell commands is not tracked, and Anthropic says to use version control for those.
- **Read every diff.** Anthropic's security page says you are responsible for reviewing proposed code and commands before you approve them. AI-written code can run and still be wrong: see [fix bugs in AI-generated code](https://www.masaischool.com/resources/build/fix-bugs-in-ai-generated-code).
- **Know your permission mode.** Anthropic's docs say that on version 2.1.283 or later a new terminal or VS Code session starts in auto mode, where a separate classifier reviews risky actions instead of you. On earlier versions that applied only to Pro, Max, Ultra and Team plans. Manual mode asks before edits and commands, and Shift+Tab switches modes. While you are learning, Manual mode lets you see every action.
- **Keep --dangerously-skip-permissions for isolated machines.** The docs list bypassPermissions mode as for containers and virtual machines only.
- **Keep secrets out of reach.** Never paste API keys into prompts or commit .env files. A deny rule such as Read(./.env) blocks Claude's file tools from reading that file. The [secure a vibe-coded app](https://www.masaischool.com/resources/build/secure-a-vibe-coded-app) checklist covers more.
- **Treat outside text as untrusted.** Web pages, issues and MCP tool results can contain instructions aimed at the model. Anthropic's docs say to verify you trust each MCP server, and note that curl and wget are not auto-approved by default in Manual mode.
- **Mind the cost.** Check /usage, and run /clear between unrelated tasks. Anthropic's best-practices page says this stops the next task paying for the previous task's history.

## What Claude Code cannot do, and free or low-cost alternatives

### Limits

- It is not on the free claude.ai plan, it needs an internet connection and it is limited to Anthropic's supported countries.
- It can be wrong. It runs on an LLM, so it can write a plausible fix that fails or breaks something else. Tests and your own review stay necessary. [What is a large language model](https://www.masaischool.com/resources/what-is-a-large-language-model) explains why.
- It does not replace knowing how to read code. If you cannot tell whether a diff is right, you cannot supervise the tool.
- A long session fills its context window. Anthropic's docs say a conversation is compacted automatically before the window fills, and a summary can lose detail, so start unrelated tasks fresh.
- It cannot certify that its code is secure. Check with [free testing and security tools for AI code](https://www.masaischool.com/resources/tools/free-testing-and-security-tools-for-ai-code).

### Alternatives, with free status read on 11 November 2026

| Tool | What it is | Free or low-cost status | Catch |
|---|---|---|---|
| OpenCode | An MIT-licensed open-source coding agent for terminal, desktop and more. It supports 200+ providers through Models.dev, including local models | Its site says a set of free models needs no account. OpenCode Go starts at $10 a month. You can also sign in with ChatGPT Plus or Pro, or GitHub Copilot | The free model list is chosen by OpenCode and can change |
| Antigravity CLI | Google's terminal agent. Google moved free individual access here from Gemini CLI | The Individuals plan is $0 with 'basic weekly rate limits'. Gemini CLI stopped serving free individual users on 18 June 2026 | The weekly limits are not published. Gemini CLI remains for enterprise licences and paid API keys |
| Codex CLI | OpenAI's terminal agent, with its code on GitHub | OpenAI's pricing page lists Codex as included in ChatGPT Free, Go, Plus and Pro. An API key also works, billed at API rates | Free-plan limits are not published, and the CLI's README recommends signing in with Plus or higher |
| Cline | An open-source agent as an extension for VS Code, JetBrains, Cursor and Windsurf, plus a desktop app and a CLI | Free to install. It has no models of its own, so you bring an API key or a local model | Your cost is whatever model you connect |
| Ollama with an open model | Runs open-weights models locally. Its library pages show ollama launch commands for Claude Code and OpenCode | Free apart from your hardware. Gemma 4 tags start at about 4.6 GB | Local models are usually weaker, and tool support varies |

A simple way to choose: if you already pay for ChatGPT, try Codex. If you have a Google account, try Antigravity's free plan. If you want freedom from one vendor and can handle setup, try OpenCode or Cline with a free API tier. The lists of [free AI coding agents](https://www.masaischool.com/resources/tools/free-ai-coding-agents) and [free AI coding assistants](https://www.masaischool.com/resources/tools/free-ai-coding-assistants) track current plans, and [run a free local coding assistant](https://www.masaischool.com/resources/build/run-a-free-local-coding-assistant) covers the Ollama route.

## Frequently asked questions

### Is Claude Code free?

No. As of 11 November 2026 Anthropic's setup docs say the free claude.ai plan does not include Claude Code. You need Pro ($17 a month billed annually, or $20 monthly), Max, another paid plan, or Claude Console with pre-paid API credits. Free options exist elsewhere, such as Antigravity's free plan, OpenCode's free models and local models through Ollama.

### Do I need an API key to use Claude Code?

Not necessarily. You can sign in with a Pro, Max, Ultra, Team or Enterprise account, which uses your subscription, or use Console API credits billed per token. If you set the ANTHROPIC_API_KEY environment variable and approve it when asked, Claude Code skips the login prompt. Keep keys out of your repository.

### What is the difference between Claude Code and Claude chat?

Claude chat answers in a conversation. Claude Code works inside your project: it reads and edits files and runs commands, asking for approval depending on its permission mode. Claude Code needs a paid plan or API credits, while chat has a free plan.

### Is Claude Code an AI agent?

Yes, in the sense that it runs a loop: it decides on an action, uses a tool such as editing a file or running a test, reads the result and repeats. Anthropic describes it as an agentic coding tool. 'Claude agent' is a looser search term, and the AI agent guide on this site explains the general idea.

### Is Claude Code better than Cursor?

They are different kinds of tools, so 'better' depends on how you work. Cursor is an editor with AI built in. Claude Code is an agent you can run in a terminal or as an extension inside VS Code or Cursor. Try both on the same small task and compare the diffs, the cost and how much you had to correct.

### Can I use Claude Code offline or with other models?

Not offline: the docs list an internet connection as a requirement. Claude Code also works with Amazon Bedrock, Google Cloud's Agent Platform and Microsoft Foundry, and Ollama documents a setup that points it at local open-weights models. Local models are usually less capable than Claude models, so expect weaker results.

### How do I install Claude Code on Windows?

Open PowerShell and run irm https://claude.ai/install.ps1 | iex, or use WinGet with winget install Anthropic.ClaudeCode. Open a new terminal, run claude --version, then start claude in a project folder. Git for Windows is recommended so the Bash tool works. WSL 2 is the route if you want sandboxing, which native Windows does not support.

## Sources checked

- [Surfaces, install commands, what Claude Code does, Claude Academy](https://code.claude.com/docs/en/overview)
- [System requirements and the account requirement: free claude.ai plan excluded](https://code.claude.com/docs/en/setup)
- [Free, Pro and Max prices; Claude Code in Pro and Max, not Free](https://claude.com/pricing)
- [First-session steps, login options, Console credits](https://code.claude.com/docs/en/quickstart)
- [Enterprise cost averages, /usage, /clear](https://code.claude.com/docs/en/costs)
- [Auto mode default from v2.1.283, bypassPermissions for isolated environments](https://code.claude.com/docs/en/permission-modes)
- [Manual mode prompts, Read(./.env) deny rule](https://code.claude.com/docs/en/permissions)
- [User responsibility, prompt injection protections, network command approval](https://code.claude.com/docs/en/security)
- [/rewind and the Bash-changes limitation](https://code.claude.com/docs/en/checkpointing)
- [Specific prompts, /clear between tasks, Esc to interrupt](https://code.claude.com/docs/en/best-practices)
- [Pointing Claude Code at Ollama models](https://docs.ollama.com/integrations/claude-code)
- [OpenCode: MIT, free models, Go plan](https://opencode.ai/)
- [Gemini CLI free individual access ended 18 June 2026](https://developers.googleblog.com/an-important-update-transitioning-gemini-cli-to-antigravity-cli/)
- [Antigravity Individuals plan $0, basic weekly rate limits](https://antigravity.google/pricing)
- [Codex included in ChatGPT Free, Go, Plus and Pro; API key option](https://learn.chatgpt.com/docs/pricing)
- [Cline: open source, bring your own model](https://cline.bot/)
