What Is Claude Code? Install, Pricing, Commands (2026)

What it is: Claude Code is Anthropic’s AI-powered coding assistant. It runs in your terminal, IDE, desktop app, Slack workspace, web browser, or CI/CD pipeline, reads your real codebase, edits files with your approval, runs commands, manages git, and handles multi-step tasks autonomously.
Who it is for: Software developers, engineers, technical founders, and curious learners who want an AI assistant that operates inside their code rather than alongside it.
Best if: You want depth over speed, full project context, and agentic workflows that can ship code without constant copy-paste.
Skip if: You only need inline autocomplete (see our Copilot comparison) or you prefer a paid GUI IDE. Get one issue per day of what just shipped in AI in our free daily newsletter.

Pairing tip: Claude Code can drive Playwright (Microsoft’s browser-automation library) through the official Playwright MCP server — the most efficient agentic browser architecture in 2026. See Playwright vs Claude Computer Use for the full comparison and the hybrid setup.

Related Claude capability: Beyond the terminal, Claude can also operate a real desktop or browser via Computer Use and Browser Use. See Claude Computer Use Best Practices for the production guide based on Anthropic’s May 2026 blog post.

What is the bottom line on Claude Code?

Claude Code is Anthropic’s AI coding assistant that operates inside your actual codebase rather than alongside it. You open a terminal in your project folder, type claude, and the assistant reads your files, understands the architecture, and can edit, debug, refactor, test, and commit code with your approval. It costs $17–$20/month bundled with Claude Pro, runs on Mac, Windows, Linux, in VS Code and JetBrains IDEs, in Slack, on the web, and in CI/CD via GitHub Actions and GitLab. Under the hood it uses Claude Opus 4.7 or Sonnet 4.6, and as of May 2026 it leads the Terminal-Bench 2.0 agentic coding benchmark.

What are the key takeaways?

• Install in 30 seconds with a single shell command on Mac, Linux, WSL, or Windows.
• Included in Claude Pro ($17/month annual, $20 monthly). Higher quotas on Max, Team, and Enterprise plans.
• Reads your real codebase — no copy-paste, no manual context loading.
• Agentic by design — can plan, execute, test, and commit multi-step changes with your approval.
• Available everywhere: terminal, VS Code, JetBrains, Cursor/Windsurf, Slack, web, iOS (preview), GitHub Actions, GitLab CI.
• Three auth options: Claude subscription, Anthropic Console (pay-as-you-go API), or cloud providers (Bedrock, Vertex AI, Foundry).
• Built on Claude Opus 4.7 / Sonnet 4.6 — the same models that lead public coding benchmarks.

What is Claude Code exactly?

Claude Code is an AI coding assistant from Anthropic. In Anthropic’s framing: “Work with Claude directly in your codebase. Build, debug, and ship from your terminal, IDE, Slack, or the web.”

Unlike traditional AI coding tools that you talk to in a chat window and then copy code back into your editor, Claude Code operates inside your project. It reads files, edits them with permission, runs shell commands, runs tests, makes git commits, and can chain multi-step tasks together. It’s a CLI-first tool that also has first-class integrations with most IDEs and collaboration platforms developers actually use.

How is Claude Code different from ChatGPT, Cursor, and GitHub Copilot?

Tool	Primary mode	Reads your codebase?	Edits files autonomously?	Starting price
Claude Code	Terminal + IDE + Slack + web	Yes — full project	Yes, with permission gates	$17–$20/mo (Claude Pro)
ChatGPT	Chat window	No — you paste	No	$20/mo (Plus)
Cursor	VS Code fork (GUI)	Yes	Yes, in editor	$20/mo (Pro)
GitHub Copilot	IDE autocomplete + chat	Partial (active file)	Limited (agents preview)	$10/mo (Individual)
OpenAI Codex CLI	Terminal	Yes	Yes	$20/mo (ChatGPT Plus)

The key differences: Claude Code is the only one that’s terminal-first AND has full IDE integrations AND has Slack/CI/CD plugins. For deeper feature-by-feature breakdowns see Claude Code vs Cursor vs Copilot and Codex vs Claude Code.

How do you install Claude Code on Mac?

One command in your terminal:

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

Or via Homebrew:

brew install --cask claude-code

The native install auto-updates in the background. The Homebrew install does not — run brew upgrade claude-code periodically. After install, run claude in any project directory and you’ll be prompted to log in.

How do you install Claude Code on Windows?

PowerShell (recommended):

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

Windows CMD:

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

WinGet:

winget install Anthropic.ClaudeCode

On native Windows, install Git for Windows first so Claude Code can use the Bash tool. Without it, Claude Code falls back to PowerShell for shell commands, which works but limits some workflows. WSL installations don’t need Git for Windows.

If you see the error The token '&&' is not a valid statement separator, you’re in PowerShell, not CMD — use the PowerShell command above. If you see 'irm' is not recognized, you’re in CMD — use the CMD command.

How do you install Claude Code on Linux?

Same one-line install as Mac:

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

Native package managers are also supported on Debian/Ubuntu (apt), Fedora/RHEL (dnf), and Alpine (apk). See the official setup docs for the package-manager commands. WSL on Windows uses the same Linux install path.

How much does Claude Code cost?

Plan	Price	Claude Code included?	Best for
Claude Pro	$17/mo annual ($200 upfront) or $20/mo monthly	Yes	Individual developers
Claude Max	Higher tier (see Anthropic pricing)	Yes, higher quotas	Heavy users running agent teams
Claude Team	Per-seat (see Anthropic pricing)	Yes	Small engineering teams
Claude Enterprise	Custom	Yes	Large orgs with SSO and admin needs
Anthropic Console (API)	Pay-as-you-go: $5 / $25 per M tokens (Opus 4.6)	Yes — metered usage	Teams wanting cost transparency

The simplest path for most developers: Claude Pro for $17–$20/month. It includes generous Claude Code usage at the Sonnet 4.6 + Opus 4.7 tier. If you start hitting limits, upgrade to Max. If you want strict cost tracking, use the Console (API) path — first login auto-creates a “Claude Code” workspace for centralized cost tracking.

How do you authenticate Claude Code?

Three authentication routes, in order of simplicity:

1. Claude subscription (Pro, Max, Team, Enterprise) — recommended for individual developers and most teams. Run claude and follow the browser-based login. Credentials are stored locally; you won’t need to log in again on that machine.
2. Anthropic Console (API) — for pay-as-you-go API billing. Gives precise cost tracking and is the right path for organizations that want metered usage. Sign up at console.anthropic.com.
3. Cloud provider — Amazon Bedrock, Google Vertex AI, or Microsoft Foundry. Right path if your organization already routes AI spend through one of these providers.

Switch accounts with /login inside Claude Code at any time.

What can you actually do with Claude Code?

The headline use cases:

• Explore a new codebase. what does this project do? / where is the main entry point?
• Implement features in natural language. add input validation to the user registration form
• Fix bugs. there's a bug where users can submit empty forms — fix it
• Refactor at scale. refactor the auth module to use async/await instead of callbacks
• Write tests. write unit tests for the calculator functions
• Code review. review my changes and suggest improvements
• Git workflows. what files have I changed? / commit my changes with a descriptive message
• Documentation. update the README with installation instructions
• Multi-step builds. Plan, scaffold, implement, test, commit — all in one prompt.

Claude Code always asks for permission before modifying files, so the agentic workflow stays under your control. You can also enable “Accept all” mode for a session if you trust the changes.

What are the essential Claude Code commands?

Command	What it does
claude	Start interactive mode in current directory
claude "task"	Run a one-time task non-interactively
claude -p "query"	Run one-off query, print result, exit
claude -c	Continue most recent conversation in this directory
claude -r	Resume a previous conversation by ID
/clear	Clear conversation history (in-session)
/help	Show all available commands
/login	Switch accounts
/resume	Resume a paused conversation
Shift + Tab	Cycle permission modes
Ctrl + D / exit	Exit Claude Code

The full CLI reference is at code.claude.com/docs/cli-reference.

What is the Claude Agent SDK?

The Claude Agent SDK is the programmatic interface to Claude Code. While the terminal claude command is the interactive experience, the SDK lets you run the same agent capabilities from your own programs — building custom agents, embedding Claude Code into your CI pipelines, or composing multi-agent workflows.

Practical uses include automated PR review bots, scheduled code-quality audits, and headless task runners that triage incoming bug reports. Anthropic launched a separate “Agent SDK credits” tier in May 2026 so subscription users can allocate part of their Claude Pro/Max usage specifically toward programmatic Claude Code usage — including third-party agent frameworks like OpenClaw.

How does Claude Code use MCP (Model Context Protocol)?

MCP is an open standard from Anthropic that lets AI assistants connect to external tools, data sources, and services through a single uniform interface. Claude Code is MCP-native — you can plug in MCP servers for GitHub, Notion, Linear, Slack, your own internal APIs, or essentially any service that exposes an MCP-compatible interface.

In practice this means you can say create a Linear ticket for this bug and assign it to me or pull the last 5 PRs that touched the auth module and Claude Code will use the appropriate MCP server to complete the task without you switching context. The Model Context Protocol is documented at modelcontextprotocol.io.

What are Claude Code Skills?

Skills are reusable, named recipes that Claude Code can invoke. Think of them as “named workflows you teach the agent once and call by name later.” Examples: a release-prep skill that bumps the version, updates the changelog, tags the commit, and opens a PR; a add-tests skill that scaffolds test files matching your project’s conventions; a security-audit skill that runs your dependency scanner, summarizes findings, and proposes fixes.

Skills can be project-local (in a .claude/skills/ directory) or shared across all your projects (in your global Claude Code config). For a deeper guide see our Claude Skills explainer.

Where can you use Claude Code beyond the terminal?

• VS Code and forks (Cursor, Windsurf) — first-class extension with inline diffs and chat panel.
• JetBrains IDEs (IntelliJ, PyCharm, WebStorm, GoLand, etc.) — official plugin.
• Slack — install via the Slack App Marketplace; mention Claude Code in any channel.
• Web — claude.ai/code in any browser.
• iOS — research preview as of 2026.
• Desktop app — standalone Claude desktop on macOS and Windows.
• GitHub Actions — CI-native workflow for code review, test generation, and auto-fixing.
• GitLab CI/CD — equivalent pipeline integration.

How safe is Claude Code with my codebase?

Claude Code is designed with three safety layers:

• Permission gates by default. Every file edit, every shell command, every git operation requires explicit approval. You can also batch-approve with “Accept all” within a session.
• Permission modes you can cycle with Shift+Tab. Use restrictive mode for sensitive directories, plan mode for read-only analysis.
• The --dangerously-skip-permissions flag exists. Anthropic specifically named it that way to discourage casual use. Reserve it for ephemeral sandboxes you control end-to-end.

For production codebases, always run Claude Code on a feature branch, review every diff before merge, and never grant unrestricted shell access in environments with production credentials in scope. The CLAUDE.md file in your repo root is the right place to encode safety rules Claude Code should always follow.

What’s the best way to learn Claude Code?

1. Install it and run claude in a non-critical project. Ask it to explain the codebase.
2. Use it for real work. The fastest path to fluency is shipping one feature with it. Pick a small task you’d do anyway.
3. Read the advanced workflows guide. CLAUDE.md design, custom slash commands, agent teams, parallel workflows.
4. Study comparisons. Our Claude Code vs Cursor vs Copilot piece and Codex vs Claude Code piece show where each tool fits.
5. Stay current. Subscribe to the free daily AI brief — Claude Code ships meaningful updates every couple weeks, and most users miss half of them.

Frequently asked questions

Is Claude Code free?
No, but it’s included in Claude Pro starting at $17/month annual or $20 monthly. There’s no separate Claude Code license fee on top of your Claude subscription.

Does Claude Code work offline?
No. It requires an internet connection to call the Claude API.

Which Claude model does Claude Code use?
By default it uses the latest Claude Sonnet 4.6 and Opus 4.7 models. Anthropic routes requests between them automatically based on task complexity and your plan. You can pin a specific model with command-line flags.

Can I use Claude Code with my own API key?
Yes. Sign in via the Anthropic Console (console.anthropic.com) on first login. Usage is then billed to your Console account at pay-as-you-go API rates.

How does Claude Code compare to OpenAI Codex CLI?
Both are terminal-first agentic coding tools. Claude Code is more polished on multi-file refactors and has deeper IDE integrations. Codex CLI is faster on individual tasks and ships with ChatGPT Plus. See our head-to-head comparison.

Can Claude Code access the internet during a session?
Yes, via the WebFetch and WebSearch tools (with permission). It can fetch documentation, look up package versions, and read public web pages while working on your code.

What is CLAUDE.md and do I need one?
CLAUDE.md is a project-level instruction file Claude Code reads at the start of every session. It’s where you encode rules (“never push to main without tests”), preferences (“use TypeScript strict mode”), and project context. You don’t strictly need one, but adding it is the single highest-leverage thing you can do to make Claude Code’s output match your team’s standards.

Does Claude Code read my .env file?
Only if you explicitly let it. The default permission gates apply to .env just like any other file. Many teams add .env to a CLAUDE.md “never read” list as an extra safeguard.

What is the Claude Code Router?
An open-source project that lets Claude Code route requests across multiple model providers (Anthropic, OpenAI, local Ollama, etc.) based on cost and task type. Not built by Anthropic. Useful for cost optimization on heavy workloads.

Can I run Claude Code in GitHub Actions?
Yes — there’s an official action that runs Claude Code in CI for code review, test generation, and auto-fix workflows. GitLab CI/CD has an equivalent.

How do I cancel my Claude Pro subscription?
From claude.com, go to Settings → Billing → Cancel subscription. Claude Code stops working when the subscription ends unless you’ve also configured Console API auth.

Is Claude Code the same as Claude Cowork?
No. Claude Code is the developer-focused product for working in real codebases. Claude Cowork is the knowledge-worker-focused agent that operates on documents, files, and apps. They share underlying capability but target different jobs.

Get every Claude Code update in one daily email

Claude Code ships meaningful updates every couple weeks — new commands, new SDK features, new integrations, new pricing tiers. The Beginners in AI newsletter ships one issue every day covering what just shipped across AI, with extra weight on Claude. Free, daily, no fluff. Subscribe below.

Sources

• Claude Code product page — Anthropic
• Claude Code quickstart — official documentation
• Claude Code CLI reference
• Claude Code setup — install on every platform
• Claude Opus 4.6 announcement (Anthropic, Feb 5 2026)
• Model Context Protocol — official specification
• Claude pricing — current tiers and quotas

Related news

• Glossary: What Are Managed Agents?
• Microsoft is canceling Claude Code licenses for thousands of engineers — news, May 14–15 2026.
• Claude Mythos cracked Apple’s Mac M5 security in 5 days — news, May 14 2026.