## Definition An **agentic CLI** is a terminal-based interface to an LLM that can read and write files, run shell commands, call external tools, and coordinate sub-agents — all under a permission policy. Examples in 2026: **Claude Code**, **OpenCode**, **Cursor agent mode**, **Continue**. ## Why the CLI Won - **Composability.** Agents call shell tools; the shell is where shells live. - **Scriptability.** A CLI runs in CI, can be piped, looped, and version-controlled. Chat windows can't. - **Single permission surface.** Filesystem, git, package managers, databases, browsers, and the agent — all in one process tree with one policy to reason about. ## Settings Hierarchy (Claude Code shape) | Scope | Path | When to use | | ------- | ------------------------------------- | -------------------------------------- | | User | `~/.claude/settings.json` | Personal defaults | | Project | `.claude/settings.json` (committed) | Team conventions | | Local | `.claude/settings.local.json` (gitignored) | Per-developer overrides | ## Common Slash Commands `/init`, `/plan`, `/diagnostics`, `/cost`, `/compact`, `/resume`, `/mcp`. ## Related - [[Plan Mode and Build Mode]] - [[AGENTS.md Convention File]] - [[Permission Policy]] - [[Extension Trifecta]] - [[Claude Code Documentation]]