# Claude Code Claude Code is an agentic coding CLI tool created by [[Anthropic]]. It understands codebases, edits files, runs commands, and manages git workflows autonomously from the terminal. ## Getting started - `/powerup` — built-in interactive tutorial with 10 short lessons and live demos (added April 2026; update CLI with `claude update`) - [[Claude Code Installation]] — prerequisites, install, update, Windows/WSL setup - [[Claude Code Configuration]] — settings, scopes, environment variables - `/checkup` — audits your setup and proposes cleanups (unused skills/MCPs/plugins, bloated CLAUDE.md, slow hooks, stale CLI). See [[Claude Code Interactive Mode]] - `/skill-doctor` — what each skill costs in context and how often it fires. See [[Claude Code Skills]] ## How to use - [[Claude Code CLI Reference]] — CLI commands, flags, non-interactive usage - [[Claude Code Interactive Mode]] — slash commands, keyboard shortcuts, modes, prompt suggestions - [[Claude Code IDE Integration]] — Visual Studio Code, JetBrains IDEs - [[Claude Code Voice Mode]] — hands-free and push-to-talk voice input ## Core systems - [[Claude Code Tools]] — built-in tools (Read, Write, Edit, Bash, Agent, Grep, etc.) - [[Claude Code Memory]] — CLAUDE.md files, auto memory, project rules, imports. Since v2.1.277 (2026-09-18), Claude Code also reads `AGENTS.md` when no `CLAUDE.md` exists. See [[AGENTS.md (File Convention)]] - [[Claude Code Permissions]] — permission model, allow/deny rules - [[Claude Code Auto Mode]] — background classifier replaces manual permission prompts (research preview) - [[Claude Code Hooks]] — lifecycle hooks for automation - [[Claude Code MCP Integration]] — Model Context Protocol server support - [[Claude Code Prompt Caching]] — 1h vs 5m cache heuristics, telemetry interaction, env vars ## Extensibility - [[Claude Code Skills]] — reusable instruction packages (supersedes custom commands) - [[Claude Code Custom Commands]] — legacy slash command system - [[Claude Code Plugins]] — plugin system, marketplaces, discovery - [[Claude Marketplace]]: Anthropic's storefront at claude.com/marketplace (connectors and plugins to add, partner agents to buy, service partners to hire). Different from the git-based plugin marketplaces you add with `/plugin marketplace add` - [[Claude Code Plugin Evals]] — `claude plugin eval`: score a plugin against test cases, with and without the plugin loaded - [[Claude Code Mods]]: plugins made of JS/TS handlers that run inside Claude Code, like middleware. They can watch, rewrite or answer tool calls and prompts, add instant slash commands, and draw their own UI. On by default since v2.1.287 (launched 2026-10-01). You can ask Claude to write one for you - [[Claude Code SDK]] — programmatic integration (TypeScript, Python, CLI) ## Features - [[Claude Code Projects]] — one conversation per project; Claude splits the work into parallel cloud threads and keeps going after you leave (beta, September 2026) - [[Claude Code Tasks]] — task tracking, scheduled tasks, /loop - [[Claude Code Agent View]] — one screen for all background sessions: dispatch, peek, attach (research preview, v2.1.139+) - [[Claude Code Goal Command]]: `/goal <condition>` keeps Claude working across turns until a separate evaluator model confirms the condition (v2.1.139+) - [[Claude Code Agent Teams]] — multi-instance coordination (experimental) - [[Claude Code Review]] — multi-agent code review system - [[Claude Code Security Review]] — `/security-review` slash command + GitHub Action for automated vulnerability scans on PRs - [[Claude Code Remote Control]] — continue sessions from phone or browser - [[Claude Code Web]] — browser-based cloud sessions with parallel execution, auto-fix PRs, teleport - [[Claude Code Ultraplan]] — plan in the cloud, review in browser, execute remotely or locally - [[Claude Code Web Scheduled Tasks]] — recurring cloud tasks that run without your machine (became Routines in April 2026) - [[Claude Code Routines]] — saved cloud sessions with schedule + API + GitHub triggers (supersedes Web Scheduled Tasks; research preview since 2026-04-14, `/schedule` in the CLI) - [[Claude Code Computer Use]] — control your screen, open apps, click, type, screenshot from the CLI (macOS, research preview) - [[Claude Code Channels]] — push external events (chat, webhooks, CI) into a running session - **You Should Know** (built-in plugin, 2026-10-02): a side agent that scans Claude's output for important information you might miss and shows a note above the prompt. Off by default; enable it with `/plugin enable cc-plugin-you-should-know@builtin`. It ships as a mod (see [[Claude Code Mods]]) - [[Claude Code Artifacts]] — live, interactive web pages built from your session context, shareable across your org (beta, June 2026) - **Dynamic workflows (multi-agent orchestration)** — Claude can author and run a deterministic multi-agent workflow that fans out work across subagents and verifies it. The explicit trigger word is **`ultracode`** (changed from `workflow` in June 2026). You can still say "use a workflow for this", but a bare mention of "workflow" no longer auto-kicks-off orchestration when you're clearly referring to something else; use `ultracode` to opt in explicitly. ## Effort: how much compute Claude spends [[Claude Code effort levels|Effort]] is roughly how much compute you want Claude to spend on a task. It changes how much verification, edge-case testing and own judgement Claude puts in. [[Thariq Shihipar]]'s analogy: the same task with 12 hours versus 1 hour. - Levels: `low`, `medium`, `high`, `xhigh`, `max`. Change it mid-conversation with `/effort` - On the newest models ([[Claude Fable 5.1]], [[Claude Opus 5.5]]), switching effort mid-session doesn't break the [[Claude Code Prompt Caching|prompt cache]]. [[Lydia Hallie]] (Anthropic): make sure you're on v2.1.280+ - Default is `medium` for Opus 5.5 and [[Claude Sonnet 5.5]] in Claude Code Thariq's guidance ("Spending your effort", September 2026), in short: - The less you specify, the more effort decides for you. Ask for "a personal fitness and workout tracker app": low effort gives you a log and a simple graph, max gives you something more complex, with a heat chart. Higher effort means more choices made on your behalf - Low effort is a quick way to see what Claude would do on its own. Redesigning the `/config` menu took 1 minute on low (an interactive sketch) and 28 minutes on max (a polished mockup with flow walkthroughs). He prefers low first - With a detailed spec, the results of different models and effort levels converge - Higher effort fixes edge-case failures; it doesn't fix a wrong approach. It helps most on tasks with many hidden edge cases: hardware, code review, security - His rule of thumb: low for quick in-the-loop work (brainstorming, sketching, easy changes); medium for most regular software engineering (feature work); high when verification matters (a bug in a brownfield codebase); max for fully autonomous hard problems (end-to-end build and verify, security vulnerabilities in critical software) His loop: write a spec, ask Claude to interview you about the missing details, implement on low, review and iterate on low, then verify and test on high. I like that loop: runs stay cheap and fast while you're still shaping the idea, and you pay for high effort only when checking the result. Task-level numbers are in [[Claude Opus 5.5]]. ## Recent changes (early October 2026) - **Mods launch (2026-10-01)**: [[Claude Code Mods]] are on by default from v2.1.287, and the early access flag `CLAUDE_CODE_ENABLE_FUNCTION_HOOKS` is ignored. Since the launch, `/diff` and `AGENTS.md` loading ship as built-in mods - **2026-10-02**: the You Should Know built-in plugin (see [[#Features]]) ## Recent changes (September 2026) Versions 2.1.269 to 2.1.280 shipped in eleven days: - **2.1.280 (09-22)**: [[Claude Opus 5.5]] is the default Opus model (1M context). Pro and Team Standard plans now default to Opus too, like Max. Auto mode stops retrying actions the safety check declined. The max MCP tool description length (2,048 characters) is now configurable with `CLAUDE_CODE_MAX_MCP_DESCRIPTION_LENGTH`. New VS Code dialogs: `/status`, `/sandbox`, `/chrome`, `/skills`, `/plan`, `/export` - **2.1.278 (09-19)**: the auto mode classifier runs server-side by default, with no classifier overhead charge. See [[Claude Code Auto Mode]] - **2.1.277 (09-18)**: `AGENTS.md` support (fallback when `CLAUDE.md` is missing). The `TaskOutput` tool is gone; Claude reads background output with Read - **2.1.275 (09-17)**: a send-now key (`ctrl+enter`) interrupts the turn and sends your queued messages. Skills and plugins sync from claude.ai to the terminal (opt out with `syncClaudeAiSkills: false` / `syncClaudeAiPlugins: false`) - **2.1.271 (09-14)**: `omitClaudeMd` in agent frontmatter lets custom subagents skip `CLAUDE.md` files. Fast mode works in remote sessions - **2.1.269 (09-11)**: `claude plugin eval` ([[Claude Code Plugin Evals]]), `/output-style` in remote and headless sessions, and `CLAUDE_CODE_WORKFLOW_MAX_CONCURRENT_AGENTS` (1-256) to raise the workflow concurrency cap ## Practical guides - [[Claude Code Tips and Best Practices]] — workflows, worktrees, prompt techniques, utilities - **claude.dev**: Anthropic's developer hub ("tips, tricks, and POVs from Anthropic's developers"). Articles in Agents, Engineering, Playbooks, Skills and Tutorials (getting the most out of Opus 5.5, spending your effort, Claude Code mods), plus Mods, Docs and Terminal sections and videos with founders and engineers (Ramp, DoorDash) ## Privacy See https://docs.anthropic.com/en/docs/claude-code/overview#license-and-data-usage ## References - Documentation: https://docs.anthropic.com/en/docs/agents-and-tools/claude-code/overview - Source code: https://github.com/anthropics/claude-code - Usage: https://docs.anthropic.com/en/docs/claude-code/common-tasks - CLI usage: https://docs.anthropic.com/en/docs/claude-code/cli-usage - Cheat sheet: https://cc.storyfox.cz/ - A conversation on Claude Code: https://www.youtube.com/watch?v=Yf_1w00qIKc - Claude Code Unpacked: https://ccunpacked.dev/ - Official news account on X: https://x.com/ClaudeDevs - What's new (way to keep up to date): https://code.claude.com/docs/en/whats-new - Official YouTube channel: https://www.youtube.com/@claude - Changelog: https://code.claude.com/docs/en/changelog - Hacker News discussion of AGENTS.md support (September 2026): https://news.ycombinator.com/item?id=49760187 - claude.dev (Anthropic developer hub): https://claude.dev - Thariq Shihipar, "Using Claude Code: Spending your effort" (X article, 2026-09-25; mirror): https://claude.dev/blog/spending-your-effort/ - Lydia Hallie on switching effort without breaking the prompt cache (X, 2026-09-22) - Mods overview (official docs): https://code.claude.com/docs/en/plugins/mods/overview - @ClaudeDevs on the You Should Know plugin (X, 2026-10-02) - Claude Marketplace: https://claude.com/marketplace ## Tutorials and courses - Official tutorials: https://docs.anthropic.com/en/docs/claude-code/tutorials - Free DeepLearning.AI course: https://learn.deeplearning.ai/courses/claude-code-a-highly-agentic-coding-assistant - https://www.youtube.com/watch?v=zcHY88VI1oc - Coding: https://www.youtube.com/watch?v=8YkkvVe_Z8w - Tips: https://www.youtube.com/watch?v=TiNpzxoBPz0 ## Related - [[AI Agent Harness]] - [[Claude]] - [[Claude Code via GitHub Copilot]] - [[Claude Design]] - [[Claude Opus 4.7]] - [[Claude Fable 5.1]] - [[Claude Opus 5.5]] - [[Claude Sonnet 5.5]] - [[Claude Code Mods]] - [[Claude Code Plugins]] - [[Claude Marketplace]] - [[Thariq Shihipar]] - [[Lydia Hallie]] - [[Claude Code Projects]] - [[Claude Managed Agents]] - [[Anthropic CLI (ant)]] - [[Kimi CLI]] - [[GitHub Copilot CLI]] - [[OpenClaude]] - [[Claude Cowork]] - [[AI Agents]] - [[AI Agent Skills]] - [[Claude Code Skills]] - [[Beads]] - [[Ralph TUI]] - [[Claude Epic Status Line]] - [[Claude Account Switcher]] - [[Claude companion]] - [[Conductor]] - [[Codex plugin for Claude Code]] - [[Hidden Folders Access plugin for Obsidian]] - [[Edgee]] - [[2026-10-04 Claude effort levels - Draft on low, verify on high]] - [[Claude Code effort levels]]