What is Claude Code

Claude Code is an AI-powered interactive terminal coding assistant that lets you write, edit, debug, and understand code through natural language.

Claude Code is a CLI tool built on Anthropic’s Claude AI. You run it in your terminal, describe what you want in plain English, and Claude reads your codebase, edits files, runs commands, and explains what it did — all within your existing development environment.

It is designed for software engineers who want an AI collaborator that works where they already work: in the terminal, inside their editor, alongside their existing tools.

Quickstart

Get Claude Code running and make your first request in under 5 minutes.

Installation

Install via npm, native binary, or platform package manager.

Core Concepts

Understand the tools, slash commands, skills, and agent system.

Configuration

Configure models, permissions, MCP servers, and hooks.

Key features

Interactive REPL

Claude Code runs as a persistent interactive session in your terminal. You type requests in natural language, Claude responds with actions and explanations, and the conversation continues. The session maintains context across multiple turns so you can iterate — ask Claude to fix a bug, then ask it to add tests for the fix, then ask it to update the documentation.

Built-in tool system

Claude Code has direct access to a set of built-in tools it uses autonomously to complete tasks:

• Bash — run shell commands, execute tests, manage processes
• FileRead / FileWrite / FileEdit — read, create, and modify files
• Grep — search file contents with regular expressions
• Glob — find files by name pattern
• WebFetch — fetch and read content from URLs
• WebSearch — search the web for documentation and answers
• NotebookEdit — read and edit Jupyter notebooks

Claude decides which tools to use based on your request. You control which tools are permitted through the permissions system.

Slash commands

Type / inside a session to access built-in slash commands. These control Claude Code’s behavior, configure its settings, and trigger specific workflows. Examples:

Command	Description
/help	Show all available commands
/model	Switch the AI model
/init	Generate a CLAUDE.md file for your project
/mcp	Manage MCP server connections
/hooks	View hook configurations
/skills	List available skills
/ide	Manage IDE integrations
/compact	Compact the conversation to free context
/clear	Start a new session
/exit	Exit Claude Code

Model Context Protocol (MCP)

Claude Code supports MCP, which lets you connect external tools and data sources to your session. An MCP server can expose tools, resources, or prompts — for example, a database query tool, a GitHub integration, or a custom internal API. Use /mcp to add, enable, and disable MCP servers without leaving your session.

Hooks

Hooks are shell commands that Claude Code runs automatically on tool events. You can configure a hook to run ruff format after every file edit, run your test suite before Claude exits, or log activity to a file. Hooks are deterministic and cannot be skipped by Claude — see Configuration: Hooks for setup details.

Skills

Skills are on-demand capabilities stored as Markdown files in .claude/skills/. You or Claude can invoke a skill by typing /<skill-name>. A skill file contains instructions Claude follows when invoked — useful for repeatable workflows like running a deployment, generating a session report, or following a project-specific review checklist.

Agents

Claude Code can spawn sub-agents to work on tasks in parallel. Agents are isolated Claude instances that receive a task, complete it, and return results. This lets Claude break large tasks — like migrating an API across dozens of files — into parallel workstreams. Use /agents to see the agent configuration for your session.

IDE integration

Claude Code integrates with VS Code and JetBrains IDEs. When connected, Claude can see your open file, cursor position, and selected text. Use /ide to manage the connection and install the editor extension.

Remote sessions

Claude Code supports remote sessions hosted on claude.ai. A remote session runs Claude Code in a cloud environment, accessible from any browser, with files streamed between your local machine and the remote host using the remote-control command.

Who it is for

Claude Code is built for software engineers and teams who:

• Work primarily in the terminal
• Want AI assistance that operates on real files in their real codebase
• Need automation that integrates with existing shell scripts, CI, and editors
• Build on codebases too large to paste into a chat interface

Claude Code works best when run from within a git repository. It uses the repository structure to understand project context and provide more accurate responses.