Claude Code Series Tutorial 1: What is Claude Code?

1.1 What is Claude Code?

One Sentence Definition

Claude Code is an AI programming agent (Agent) running in the terminal (Terminal) launched by Anthropic. Developers describe tasks in natural language, and it can actively read the codebase, propose modifications, execute commands, run tests, operate Git, and even request your approval before making changes — not just a "chatbot" that gives you advice, but a "partner" that can directly get things done.

Autonomous Programming Agent in the Terminal

To understand Claude Code, the key is to grasp the word "agentic." Traditional AI coding assistants are more like consultants — you ask questions, they give advice, and then you have to change the code yourself. Claude Code is an agent: you give it a goal, and it plans the steps, locates relevant files, executes operations, and can do everything the Shell can do — modify files, run commands, debug automatically, and even access or edit files in other formats through extension tools.

This autonomy is reflected in three levels:

1. File Operations: Can read, search, create, and edit across files, not limited to a single file, understanding the entire project structure.
2. Command Execution: Can run Shell commands, set up development environments, execute tests and build scripts.
3. Workflow Orchestration: Can plan multi-step tasks and execute them autonomously, e.g., "Read all TypeScript files in the current directory, refactor them to strict mode, then run checks."

What Capabilities Does It Have?

According to Anthropic's official definition and engineering team best practices, Claude Code currently has the following core capabilities:

• Code Understanding: Analyze code structure directly in the project directory, trace call relationships, understand business logic
• Code Generation and Modification: Generate new code or perform cross-file refactoring of existing code based on natural language instructions
• Command Execution: Execute Shell commands, run test, build, and deployment scripts, and interpret execution results
• Persistent Memory: Record project specifications and development preferences through the CLAUDE.md file, maintaining consistent context across sessions
• Context Compression: Automatically compress lengthy conversation history to keep context while saving tokens
• Multi-Model Switching: Support different model versions such as Sonnet, Opus, Haiku, flexibly choose based on task complexity
• Background Tasks: Support running in the background and executing asynchronous tasks

Final Summary

You can think of Claude Code as: a code-knowing colleague sitting next to you; you describe the requirements, and they directly modify the files.

It is not a "black-box code generator" that replaces you, but an "automatic pair programmer" — you still steer the direction, and it handles the details. Just like in a real pair programming session, you are in charge of review and direction, while it handles implementation and test verification, truly achieving a collaboration mode of "you describe the requirements, it does the work."