Anthropic’s agentic coding CLI and IDE extension—reads and edits files, runs commands, and operates through tool use with optional human approval gates.
Claude Code is a terminal-based agent that pairs a Claude model with file-editing, shell, and MCP tool access. It is distinct from the Claude model family: it is an agent harness, not a tier.
Why it shows up in study material
Claude Code illustrates the ReAct loop in practice: it observes code, reasons about a task, runs tools, and iterates—often with human-in-the-loop approval for risky actions.