The Complete Claude Code CLI Guide (github.com)

Anthropic’s Complete Claude Code CLI Guide documents Claude Code—a terminal-native, agentic coding assistant that reads your repo, edits files, runs shell commands, and orchestrates sub-agents via natural language. The guide covers installation (npm install -g @anthropic-ai/claude-code), session and output controls (claude, --print, --resume, --output-format json/markdown/text), model selection, sandboxing, and new 2024 features like the Skills system, Hooks, Slash Commands, MCP integration for external data, and Plugins. It also prescribes developer patterns such as a community CLAUDE.md to persist project context and speed up sessions. Technically important: Claude Code exposes fine-grained tools (Read, Write, Edit, Bash, Grep, Glob, WebFetch/WebSearch, NotebookRead/Edit, Task/TodoWrite) with an incremental permission model ("ask", "allow", "deny") configurable via .claude/settings.json and sandboxing on macOS/Linux. Read handles code, images, PDFs and notebooks; Edit requires exact-string replacements; Bash supports background runs and timeouts. Common workflows include non-interactive CI usage, JSON output for automation, and guarded git operations. For the AI/ML community this means faster, scriptable dev loops and reproducible automation—but also a need for careful permission/secret management and repo hygiene when enabling direct code execution and network access.