Claude Code Cheat Sheet

⌨ Keyboard Shortcuts

Essentials

EnterSend message / confirm EscCancel current input Esc EscRewind menu / summary Ctrl+CCancel input or generation Ctrl+DExit Claude Code Shift+TabCycle permission modes (incl. Auto) Ctrl+BSend task to the background ←←Open the Agent View NEW

Navigation & Display

Ctrl+RSearch command history Ctrl+LClear screen (keeps conversation) Ctrl+OFullscreen transcript — toggle verbose thinking Ctrl+TShow/hide the task list Ctrl+SStash the current prompt ↑/↓Navigate command history ←/→Cycle tabs / footer { / }Jump between prompts in transcript NEW

Editing & Model

\ + EnterMulti-line input Ctrl+GOpen in external editor Ctrl+VPaste image (screenshots, diagrams) Ctrl+KDelete to end of line Ctrl+UDelete from cursor to start Ctrl+YPaste deleted text Ctrl+_Undo last action Alt+B / Alt+FMove back / forward one word Option+PModel selector Option+TToggle thinking (no effect on Fable 5) NEW Alt+OToggle fast mode (Opus 4.8) ?Show all shortcuts Space (held)Push-to-talk (if /voice on) Ctrl+JNewline (control sequence) Option+EnterNewline (macOS) Ctrl+X Ctrl+KKill all background agents (×2 to confirm) Ctrl+X Ctrl+EOpen in external editor (alias Ctrl+G)

Vim Mode

vVisual mode — character selection VVisual-line mode — line selection / (NORMAL)Reverse history search (like Ctrl+R) NEW EscBack to normal mode

Quick Prefixes

/Slash command or skill !Bash mode — or background job in Agent View NEW @Autocomplete file/folder/agent

Confirmation Dialogs

Y / EnterConfirm N / EscDecline Ctrl+EToggle permission explanation Ctrl+DToggle permission debug info

Tip: slash command autocomplete now fills the command instead of running it — press Enter to run (v2.1.162). /scroll-speed tunes wheel speed NEW. /keybindings to customize (17 contexts).

⚡ Slash Commands

Session & Context

`/clear`	Reset the conversation (alias: /reset, /new)
`/compact [focus]`	Compact with optional focus
`/resume [session]`	Resume by ID, name, or PR URL (alias: /continue)
`/rewind`	Checkpoint + "Summarize up to here" NEW
`/cd <path>`	Change cwd without breaking the prompt cache NEW
`/rename [name]`	Rename the session
`/branch [name]`	Branch the conversation (alias: /fork)
`/export [file]`	Export the conversation to text
`/copy [N]`	Copy response N to the clipboard
`/cost`	Costs + tokens + per-model & cache-hit breakdown
`/context`	Colored grid of context consumption
`/tasks`	List and manage background tasks

Model & Configuration

`/model [name]`	Switch model — fable, opus, sonnet, haiku NEW
`/effort [level]`	Slider — low/medium/high/xhigh/max + ultracode NEW
`/fast [on\|off]`	Fast output — defaults to Opus 4.8 ($10/$50) NEW
`/focus`	Condensed view: last prompt + tools + answer
`/tui`	Flicker-free fullscreen rendering (alt-screen)
`/config`	Config panel (Editor mode, thinking, model fallback) (alias: /settings)
`/permissions`	Manage perms + Recent tab (retry with `r`) (alias: /allowed-tools)
`/mcp`	Manage MCP servers & OAuth
`/hooks`	View configured hooks
`/theme`	Switch / build a custom theme (~/.claude/themes/)
`/color [color]`	Prompt bar color
`/sandbox`	Toggle sandbox mode
`/scroll-speed`	Tune scroll speed (live preview) NEW

Project & Analysis

`/init`	Create CLAUDE.md for the project
`/memory`	Edit CLAUDE.md & memory, toggle auto-memory
`/add-dir <path>`	Add a working directory
`/diff`	Interactive diff, scrollable detail (`j/k`, PgUp/PgDn) NEW
`/goal`	Set a completion condition — Claude works until it holds NEW
`/code-review`	Correctness bugs (`--comment` PR, `--fix` applies) NEW
`/simplify`	Cleanup-only review (reuse, quality) + applies fixes NEW
`/security-review`	Scan for vulnerabilities
`/ultrareview`	Fleet of cloud bug-hunting agents NEW
`/workflows`	Manage dynamic workflows (multi-agent orchestration) NEW
`/insights`	HTML usage analysis report
`/stats`	Daily usage, streaks, preferences
`/ultraplan`	Cloud plan, browser review, local/remote execution
`/pr-comments [PR]`	GitHub PR comments
`/autofix-pr`	Fix a PR from the terminal or web
`/plan [desc]`	Enter plan mode
`/batch [task]`	Parallel changes across worktrees
`/debug [desc]`	Session troubleshooting
`/advisor`	Second-opinion model consulted mid-task

Extensions & Plugins

`/plugin [list]`	Manage plugins — `list --enabled/--disabled` NEW
`/reload-plugins`	Reload skills/agents/MCP without restart
`/reload-skills`	Re-scan skill directories without restart NEW
`/agents`	Running (● N) + Library tabs
`/skills`	List sorted by tokens
`/chrome`	Configure Claude in Chrome + "Select browser…" NEW
`/ide`	Manage IDE integrations

Advanced & Remote

`/loop [interval]`	Repeat — omit interval = self-pacing
`/schedule`	Scheduled cloud tasks
`/btw`	Quick question, no tools/history (`c` = copy) NEW
`/voice`	Push-to-talk voice dictation
`/teleport`	Resume a web session locally
`/desktop`	Transfer to the Desktop app
`/web-setup`	GitHub App connection for Claude Code on the web NEW
`/remote-control`	Make the session reachable from claude.ai
`/remote-env`	Configure the default remote environment

System & Account

`/help`	Help and command reference
`/doctor`	Check install, auth, last update result
`/status`	Version and account info
`/usage`	Limits + skills/subagents/plugins/MCP breakdown NEW
`/extra-usage`	Configure extra usage
`/feedback`	Report (may include 24h/7d sessions) (alias: /bug)
`/release-notes`	Interactive version picker
`/powerup`	Animated lessons to learn Claude Code
`/team-onboarding`	Generate a teammate ramp-up guide
`/login / /logout`	Sign in / Sign out
`/upgrade`	Upgrade your plan
`/privacy-settings`	Privacy settings
`/keybindings`	Open the keybindings file
`/terminal-setup`	Configure terminal keybindings
`/statusline`	Status line config + `refreshInterval`
`/install-github-app`	Install the GitHub Actions app
`/install-slack-app`	Install the Slack app via OAuth
`/exit`	Quit (alias: /quit)
`/mobile`	QR code for the mobile app (alias: /ios, /android)
`/passes`	Share a free week of Claude Code
`/stickers`	Order Claude Code stickers

🚀 CLI Launch Flags

Starting a Session

`claude`	Interactive session
`claude "query"`	Start with an initial prompt
`claude -p "query"`	Print mode — result and exit
`claude -c`	Resume the last conversation
`claude -r "id"`	Resume by name or ID
`claude -n "name"`	Name the session
`claude -w branch`	Isolated git worktree
`claude --safe-mode`	Disable ALL customizations (debug) NEW

Model & Reasoning

`--model fable`	Claude Fable 5 — long & autonomous tasks NEW
`--model claude-opus-4-8`	Opus 4.8 — new Max/API default NEW
`--model best`	Fable 5 where available, else latest Opus NEW
`--model opus[1m]`	Opus + 1M context (native on Opus 4.7/4.8)
`--effort {low..max}`	low / medium / high / xhigh / max
`--fallback-model a,b,c`	Fallback chain (max 3) — interactive too NEW
`--thinking disabled`	Disable thinking (no effect on Fable 5) NEW

Context & Directories

`--add-dir ../path`	Additional directories
`--bare`	Skip hooks, plugins, MCP, CLAUDE.md
`--verbose`	Detail of each turn
`--debug "api,hooks"`	Debug by category
`--chrome / --no-chrome`	Enable/disable Chrome
`--ide`	Auto-connect IDE on startup

Permissions

`--permission-mode plan`	Read-only mode
`--permission-mode auto`	AI classifies permissions (Pro+)
`--allowedTools`	Auto-approved tools
`--disallowedTools`	Tools removed from context
`--tools`	Restrict built-in tools
`--permission-prompt-tool`	MCP tool for permissions (headless)
`--dangerously-skip-permissions`	Equivalent to bypassPermissions

Output & Formats

`--output-format json`	Structured JSON output
`--output-format stream-json`	Real-time streaming JSON
`--input-format`	Input format (text, stream-json)
`--include-partial-messages`	Include partial streaming events
`--json-schema '{}'`	Force an output schema
`--max-turns 5`	Limit agentic turns
`--max-budget-usd 2`	Spending limit

System Prompt

`--system-prompt "text"`	Replace the system prompt
`--system-prompt-file`	Load from a file
`--append-system-prompt`	Append text to the prompt
`--append-system-prompt-file`	Append from a file

Background & Remote

`--bg --exec "cmd"`	Attachable background shell job NEW
`--remote "task"`	Remote session on claude.ai
`--remote-control`	Enable remote control
`--tmux`	Worktree + tmux pane
`--teammate-mode`	Agent teams: auto/in-process/tmux
`--from-pr 42`	PR resume: GitHub, GitLab, Bitbucket, GH Enterprise
`--fork-session`	Fork keeping the context
`--teleport`	Resume a web session

Plugins & Config

`--plugin-dir path\|file.zip`	Load plugins (folder or .zip) NEW
`--plugin-url URL`	Fetch a plugin archive for the session NEW
`--mcp-config`	Load MCP servers
`--strict-mcp-config`	Use ONLY these MCP servers
`--settings`	Custom settings JSON file
`--setting-sources`	Settings sources (user, project, local)
`--agents`	Define dynamic agents via JSON
`--agent name`	Run as an agent
`--session-id`	Specific session UUID
`--disable-slash-commands`	Disable skills and commands
`--betas`	Beta headers for API requests

CLI Subcommands

`claude agents`	Agent View — all sessions NEW
`claude agents --json`	List live sessions as JSON NEW
`claude ultrareview`	Cloud code review for CI/scripts NEW
`claude project purge`	Clean up a project's local state NEW
`claude update`	Update (announces target version)
`claude auth login/status/logout`	Authentication management
`claude mcp add/list/remove`	Manage MCP servers
`claude mcp serve`	Expose Claude Code as an MCP server
`claude plugin init <name>`	Scaffold a new plugin NEW
`claude plugin details <name>`	Inventory + projected token cost NEW
`claude plugin install/enable`	Install / enable a plugin
`claude auto-mode defaults/config`	Auto-mode rules
`claude bedrock/vertex setup`	Interactive wizards
`claude doctor`	Diagnostics + last update result

🔌 The Big 5 — Claude Code's Extension System

1. CLAUDE.md — System Prompt

`Managed`	Enterprise policy, top priority
`Project`	`./CLAUDE.md` or `./.claude/CLAUDE.md`
`User`	`~/.claude/CLAUDE.md`

Imports: @docs/standards.md (max 5 levels)

Size: Keep < 200 lines, @path to split

2. Rules — Scoped Rules

Files in .claude/rules/*.md, activated by glob:

paths: ["src/**/*.tsx"]
MUST use TanStack Query.
SHOULD prefer named exports.

Placement based on the consumer. MUST/SHOULD/MAY

3. Custom Commands — Shortcuts

Create in .claude/commands/name.md → /name

$ARGUMENTS, $0, $1 for arguments (\$ = literal $) NEW

Live context: !`git diff` (run before the prompt)

4. Skills — Complex Workflows

Directory .claude/skills/name/SKILL.md

`description`	Auto-discovery trigger (1536 chars)
`context: fork`	Isolated subagent (no pollution)
`paths: "*.ts"`	Activate by file type
`effort: high`	Force the effort level
`user-invocable`	true = /command, false = Claude only
`allowed-tools`	Restrict the skill's tools
`disallowed-tools`	Remove tools while the skill is active NEW

Progressive disclosure: SKILL.md < 500 lines, [ref](references/). /reload-skills re-scans without restart NEW.

Builtin: code-review, security-review, simplify, init, deslopify, loop, claude-api. Hide: disableBundledSkills NEW.

5. Subagents — Parallel Specialists

Create in .claude/agents/name.md

`model`	Model (inherited by default)
`tools`	Restrict tools
`isolation: worktree`	Separate git worktree
`memory: project`	Persistent memory (user/project/local)
`permissionMode`	Dedicated permission mode
`effort`	Subagent effort level

Invoke: @"name (agent)" | --agent name | SendMessage to: <id>. Case-/separator-insensitive matching NEW.

Builtin: Explore (Haiku), Plan, general-purpose, claude-code-guide, statusline-setup, code-reviewer.

6. MCP + Plugins

MCP: Integrations via .mcp.json. Tools: mcp__server__tool. OAuth RFC 9728. stdio servers receive CLAUDE_CODE_SESSION_ID, CLAUDE_PROJECT_DIR and CLAUDECODE=1 NEW.

# Add / List / Remove
claude mcp add playwright -- npx @playwright/mcp
claude mcp list   # unapproved = ⏸ Pending
claude mcp remove playwright
# Import / JSON / Exposed server
claude mcp add-from-claude-desktop
claude mcp serve

Transports: stdio | http | sse | ws

Plugins: Skills + agents + hooks + MCP/LSP + bin/ on PATH. Auto-load from .claude/skills/ NEW. defaultEnabled: false = installs without turning on.

/plugin install name@marketplace · claude plugin init <name>

🖥 Computer Use — in the CLI

Claude sees and controls your desktop: screenshots, clicks, typing, scroll, drag & drop. Available in the CLI on macOS and Windows Desktop (Linux not supported). Multi-monitor. Pro/Max. Enable in Settings > General > Computer Use. 27 tools (screenshot, left_click, type, computer_batch, etc.). request_access required + per-app allowlist.

👁 Monitor tool

Built-in tool to spawn background watchers — tail logs, watch CI, auto-fix crashes. Each stdout line = a notification in the conversation. Replaces sleep loops in Bash. Pairs with /loop self-pacing.

🛡 Permission Modes

Mode	Behavior
Default	Asks permission for each tool
acceptEdits	Auto-approves edits, prompts on build configs NEW
Plan	Read-only. Explores & understands. Writes nothing.
Auto	AI classifies. Pro+ (Sonnet 4.6 or Opus) NEW
dontAsk	Skips prompts, ignores unapproved tools
bypass	Skips all checks (containers only)

Normal → Auto-Accept → Plan → Auto

Shift+Tab to cycle (Agent View included). Auto mode no opt-in since late May. Classifier detail in Row 5.

Pro workflow: Plan Mode to explore → Normal to implement → Auto for long runs → Rewind if needed.

Pre-approving Tools

// .claude/settings.json
"allow": ["Read", "Bash(git *)", "Bash(npm run *)"]
"ask": ["Bash(npm publish *)"]
"deny": ["Bash(rm -rf *)", "*"]

Order: deny → ask → allow (first match wins)

Glob in deny: tool names accept patterns; "*" denies all NEW

June hardening: prompts before writing shell startup files (.zshenv, .bash_login), ~/.config/git/, and build configs (.npmrc, .bazelrc, .devcontainer/) NEW.

🔗 Hooks — Automation

Events (27)

Event	When
`SessionStart`	Start / resume — `reloadSkills: true` NEW
`SessionEnd`	End of session
`UserPromptSubmit`	Before Claude processes the message
`PostUserPromptSubmit`	After submit — can override `sessionTitle`
`PreToolUse`	Before a tool — can block (exit 2)
`PostToolUse`	After a tool — `duration_ms`, `continueOnBlock` NEW
`PostToolUseFailure`	After a failed tool
`MessageDisplay`	Transform/hide displayed text NEW
`PermissionRequest`	Permission dialog appears
`PermissionDenied`	After auto-mode denial (`retry: true`)
`Stop`	End of response — `additionalContext` NEW
`StopFailure`	Turn ended by API error
`SubagentStart`	A subagent starts
`SubagentStop`	Subagent ends — `additionalContext` NEW
`TeammateIdle`	A teammate is about to go idle
`TaskCreated`	Task created via TaskCreate
`TaskCompleted`	Task completed
`PreCompact`	Before compaction — `exit 2` to block
`PostCompact`	After compaction
`Notification`	Claude sends a notification
`InstructionsLoaded`	CLAUDE.md or rules loaded
`ConfigChange`	A config file changes
`FileChanged`	A watched file changes
`CwdChanged`	The cwd changes (reload env)
`WorktreeCreate / Remove`	Worktree create / remove
`Elicitation(Result)`	Input requested by an MCP server

5 Handler Types

`command`	Shell command (+ `args: string[]` exec form) NEW
`http`	POST to an endpoint
`prompt`	Single-turn LLM (Haiku)
`agent`	Multi-turn subagent
`mcp_tool`	Direct call to an MCP tool

// Exec form (args[]) + conditional filter
"PostToolUse": [{
  "matcher": "Write|Edit",
  "if": "Edit(*.ts)",
  "hooks": [{
    "type": "command",
    "command": "npx",
    "args": ["prettier", "--write", "$CLAUDE_FILE_PATHS"]
  }]
}]

Codes: 0 = success, 2 = block (stderr → Claude), other = error

if field: filters by name AND arguments. e.g. Bash(git *), mcp__server__tool

Env vars: $CLAUDE_PROJECT_DIR, $CLAUDE_FILE_PATHS, $CLAUDE_TOOL_INPUT, $CLAUDE_SESSION_ID, $CLAUDE_EFFORT NEW

New: terminalSequence (desktop notifs/bell), effort.level visible to hooks, CLAUDE_CODE_STOP_HOOK_BLOCK_CAP (loop cap) NEW

✨ Input Superpowers

Syntax	Action
`@file`	Reference files & folders in context
`@"agent"`	Mention a specific subagent
`!command`	Run a shell command. e.g. `! git status`
`Ctrl+V`	Paste screenshots, diagrams, images
`ultrathink`	Recognized keyword — one-off deeper reasoning

Pipe Input

# Analyze logs
cat error.log | claude -p "find root cause"

# PR review
gh pr diff 42 | claude -p "security review"

# Commit message
git diff --staged | claude -p "conventional commit msg"

Background & Agent View

Ctrl+B to send to the background. claude --bg --exec "pytest -x" launches an attachable job NEW.

←← opens the Agent View — sessions running / blocked / done. /resume resumes bg sessions NEW.

--remote "task" to launch an agent on claude.ai.

Multi-directory & Worktrees

claude --add-dir ../lib --add-dir ../shared
claude -w feature --tmux # worktree + tmux

worktree.baseRef (remote default vs HEAD) and worktree.bgIsolation: "none" NEW.

Built-in Agent Tools

Filesystem	`Read`, `Write`, `Edit`, `NotebookEdit`, `Glob`, `Grep`
Shell	`Bash`, `BashOutput`, `KillBash` (+ `PowerShell`)
Web	`WebSearch`, `WebFetch`
Agents & Skills	`Agent` (Task), `Skill`, `ToolSearch`, `Monitor`
Session / Meta	`TodoWrite`, `AskUserQuestion`, `ExitPlanMode`, `PushNotification`
MCP	`mcp__<server>__<tool>` (auto-prefixed)

Restrict: --tools Read,Edit,Bash(git *) · Forbid: --disallowedTools Bash

⚙ Configuration

Hierarchy (descending priority)

Enterprise	`/etc/claude-code/managed-settings.json`
CLI Args	Flags passed at launch
Local	`.claude/settings.local.json` (gitignored)
Project	`.claude/settings.json` (committed)
User	`~/.claude/settings.json` (global)

Environment Variables

`ANTHROPIC_API_KEY`	API key (overrides subscription)
`ANTHROPIC_MODEL`	Default model
`ANTHROPIC_DEFAULT_FABLE_MODEL`	Fable 5 ID (alias `fable` + auto fallback) NEW
`ANTHROPIC_DEFAULT_OPUS_MODEL`	ID resolved by the `opus` alias
`CLAUDE_CODE_SUBAGENT_MODEL`	Model for all subagents
`CLAUDE_CODE_EFFORT_LEVEL`	Effort level (low…xhigh/max)
`CLAUDE_CODE_SAFE_MODE`	Disable all customizations NEW
`CLAUDE_CODE_DISABLE_BUNDLED_SKILLS`	Hide bundled skills NEW
`CLAUDE_CODE_ENABLE_AUTO_MODE`	Auto mode on Bedrock/Vertex/Foundry NEW
`CLAUDE_CODE_SESSION_ID`	Injected into stdio MCP servers NEW
`CLAUDE_CODE_PLUGIN_PREFER_HTTPS`	Clone GitHub plugins over HTTPS NEW
`CLAUDE_CODE_STOP_HOOK_BLOCK_CAP`	Stop-hook loop cap NEW
`CLAUDE_CODE_DISABLE_ADAPTIVE_THINKING`	Fixed thinking budget (Opus 4.6/Sonnet 4.6) NEW
`ANTHROPIC_WORKSPACE_ID`	Workload identity federation scoping NEW
`MAX_THINKING_TOKENS`	`0` = off (except Fable 5)
`DISABLE_PROMPT_CACHING[_FABLE]`	Disable prompt caching (per tier) NEW
`CLAUDE_CODE_USE_BEDROCK/VERTEX`	Third-party providers
`CLAUDE_PROJECT_DIR`	Project root (injected into hooks)

Settings.json — Key Fields

`model`	Default model (saved by `/model`)
`effortLevel`	Default effort (low/medium/high/xhigh) NEW
`fallbackModel`	Fallback chain (array, max 3) NEW
`availableModels`	Model allowlist (+ `enforceAvailableModels`) NEW
`modelOverrides`	Map IDs to provider IDs NEW
`requiredMinimum/MaximumVersion`	Version gating (managed) NEW
`disableBundledSkills`	Hide bundled skills/workflows NEW
`autoMode.hard_deny`	Rules that block unconditionally NEW
`pluginSuggestionMarketplaces`	Org marketplace allowlist NEW
`worktree.baseRef / bgIsolation`	Branch base / background isolation NEW
`alwaysThinkingEnabled`	Thinking by default (`Option+T`)
`permissions.{allow,ask,deny}`	Permission rules
`hooks`	Lifecycle hooks
`theme`	Theme name or custom JSON

Diagnostics: claude --safe-mode disables CLAUDE.md, skills, MCP, and hooks — ideal to tell whether a customization triggers a fallback or bug. Git status and directory names stay included. NEW

📁 File Structure Map

Project level (`.claude/`)

CLAUDE.md — Project memory, conventions settings.json — Shared settings (git) settings.local.json — Personal settings (gitignored) commands/ — Project slash commands skills/ — Skills (+ plugin auto-load) NEW agents/ — Project subagents rules/ — File-scoped rules claude-security-guidance.md — Security plugin rules NEW .mcp.json — Project MCP config (committable)

Global level (`~/.claude/`)

CLAUDE.md — Applies to ALL projects settings.json — Global settings commands/ — Personal commands skills/ — Personal skills agents/ — Personal agents rules/ — Global rules themes/ — Custom JSON themes plans/ — /plan & /ultraplan files keybindings.json — Custom shortcuts

Hierarchy (5 levels)

Policy > Organization > User global > User local > Project

managed-settings.json can enforce requiredMinimumVersion / requiredMaximumVersion NEW.

⏪ Rewind & Checkpoints

Esc EscOpen the rewind menu anytime /rewindSame function via command

Rewind Options

Option	Effect
Conversation	Go back in the chat. Code stays.
Code	Restore files. Conversation stays.
Full Rewind	Restore conversation AND code.
Summarize up to here	Compress earlier context, keep recent turns NEW

Limits: Side effects (Bash, deleted files, installed packages) can NOT be undone. Checkpoints only track Claude's file edits. Use Git.

Context Management

Startup budget ~6K tokens: system prompt, memory, env, MCP, CLAUDE.md, rules, skill descriptions.

Compaction: Auto at ~95%. Manual /compact [focus]. The model now preserves sensitive user instructions. CLAUDE.md re-read from disk.

"Document & Clear" workflow: Save progress → /clear → Restart. Claude reads the summary.

Save context: Scoped rules, subagents, --bare, context: fork, opus[1m] or fable (native 1M).

1M context: Fable 5, Opus 4.8/4.7, and Sonnet 4.6. Opus auto-upgraded on Max/Team/Enterprise. /usage breaks down skills/subagents/plugins/MCP NEW.

🤖 Auto Mode

Permission classifier: the AI decides whether each tool is safe or risky — no more dialogs on every Bash.

Availability NEW

Pro	GA — Sonnet 4.6 or Opus (Week 21)
Max	GA with Opus 4.7 / 4.8
Bedrock/Vertex/Foundry	`CLAUDE_CODE_ENABLE_AUTO_MODE=1` (Opus 4.7/4.8)

Activation

Shift+TabCycle to the "Auto" pill (no opt-in) NEW --permission-mode autoAt launch /configSet it persistently

Classifier — 2 Paths

Fast path: Haiku scans "safe / risky / uncertain". Thinking path: handles the uncertain cases to cut false positives. The dialog explains when a permissions.ask rule caused the prompt NEW.

Hard deny NEW

// settings.json — blocks unconditionally
"autoMode": {
  "hard_deny": ["Bash(curl * | sh)"]
}

Rules: claude auto-mode defaults · effective config: claude auto-mode config. PermissionDenied hook + {"retry": true} to react to a denial.

🧠 Models & Effort Levels

Available Models

Model	ID / alias	Notes
Fable 5 SUSPENDED	`fable`	Mythos-class, 1M ctx, 128K out, $10/$50 NEW — disabled June 12 (US gov. order)
Opus 4.8	`claude-opus-4-8`	Default Max/API, 1M ctx NEW
Opus 4.7	`claude-opus-4-7`	Default on AWS, xhigh by default
Sonnet 4.6	`claude-sonnet-4-6`	Default Pro / Team Standard
Haiku 4.5	`claude-haiku-4-5`	Ultra-fast, classifier, background

Aliases: best (Fable if available, else Opus), fable, opus, sonnet, haiku, opus[1m], opusplan, default. The best alias now resolves to Opus 4.8 while Fable 5 is suspended.

Fable 5 — automatic fallback NEW SUSPENDED

⚠ Fable 5 suspended: Disabled globally on June 12, 2026 by US government order. The fallback mechanism below is documented for reference and will apply again if/when access is restored.

Safety classifiers (cybersecurity, biology): a flagged request reroutes to Opus (4.8 on the API, 4.7 on AWS) with a notice. Thinking always on, can't be disabled. /config → "ask before switching" to decide each time.

Effort Levels

low medium high xhigh max ultracode

`high`	Default Fable 5, Opus 4.8, Sonnet 4.6
`xhigh`	Default Opus 4.7 — deep reasoning
`max`	Maximum budget, session-only
`ultracode`	Claude Code setting: xhigh + dynamic workflows NEW

Invoke

/effortInteractive slider (← →), "Faster/Smarter" labels --effort xhighAt launch "ultrathink"One-off effort in the prompt Option+TToggle thinking

🌍 Platforms

CLI

macOS	Native installer (npm or `curl`)
Linux	Native installer — PID namespace isolation
Windows	No more Git Bash — PowerShell by default NEW

Agent View NEW

claude agents (or ←←): one screen for every session — running, blocked, done. Per-session PR, elapsed duration, ! for a background job. claude agents --json for scripting.

Desktop app

Multi-session, integrated terminal, file editor + HTML/PDF preview, diff viewer. /desktop transfers a session from the CLI. Computer Use on macOS & Windows.

Web — claude.ai/code

Persistent cloud sessions, redesigned sidebar, drag-and-drop. Routines: templated cloud agents (schedule, GitHub event, API) NEW. Mobile push notifications NEW. /teleport to resume locally.

IDE

VS Code	Multiple agentic extensions simultaneously
JetBrains	Beta plugin — improved diff viewer
Devin Desktop	Formerly Windsurf (renamed June) NEW

Custom themes: build and share a palette from /theme or a plugin. Session recap: a summary of what happened while the terminal was unfocused.

🗂 Agent View — one screen for everything NEW

Research Preview (v2.1.139). A single screen lists every Claude Code session: what's running, what's blocked on you, what's done.

Open

claude agentsLaunch the Agent View ←←From a session (sidebar) claude agents --jsonScriptable output claude agents --cwd pathScope to a directory

In the view

Shift+TabCycle modes (auto included) ! <cmd>Attachable background shell job TabTitle = number of waiting sessions

`claude agents` flags

`--model / --effort`	Model & effort
`--permission-mode`	Permission mode
`--add-dir / --settings`	Context & config
`--mcp-config`	MCP servers
`--plugin-dir`	Plugins

Background sessions

"Pinned": stay alive when idle, restart in place for updates. /resume lists bg sessions with their duration (e.g. "3h 2m 5s"). Model & effort preserved on wake.

Combo: claude --bg --exec "npm test" runs a test in the background, visible in the Agent View, attachable/detachable at will.

🏗 Dynamic Workflows & Security NEW

Dynamic workflows (research preview)

A workflow is an orchestration script Claude writes for your task and runs across dozens to hundreds of subagents in the background. For tasks too large for one conversation: codebase-wide audit, large migration, cross-checked research.

# Describe the task, ask for a workflow
> create a workflow that migrates every
  internal fetch() call to the new HttpClient

Manage runs: /workflows. Trigger via the ultracode effort (xhigh + auto orchestration) or the ultracode keyword (was workflow, renamed June).

Security guidance plugin NEW

The security-guidance plugin reviews Claude's changes for vulnerabilities and fixes them in the same session:

Fast pattern check on each edit
Model review at the end of each turn
Deeper agentic review on commit / push

> /plugin install security-guidance@claude-plugins-official
> /reload-plugins

Project rules in .claude/claude-security-guidance.md.

Ultrareview NEW

/ultrareview (or claude ultrareview in CI): a fleet of bug-hunting agents runs in the cloud, findings land back in the CLI or Desktop automatically.

🎯 Pro Workflow — Get the Most Out of Claude Code

Start a Project

1.cd project && claude 2./init to generate CLAUDE.md 3.Edit CLAUDE.md → conventions & commands 4./powerup to discover features 5.Code!

Plan → Execute Loop

Shift+Tab × 2 → Plan Mode → Describe the task

Approve → Shift+Tab → Normal or Auto → "implement the plan"

/goal to set a completion condition: Claude works across turns until it holds. /ultraplan to delegate the plan to the cloud. NEW

Manage Context

/compact [focus]Compact with a focus area /clearUnrelated tasks? Reset fableLong & autonomous tasks NEW /focusCondensed view (quick read)

Watchers & Automation

MonitorBackground watcher, tail logs /loopSelf-pacing (omit the interval) /workflowsMulti-agent orchestration NEW

Prompting Techniques

Precisecat error.log | claude -p "root cause" Outcome-firstOn Fable 5: describe the result, not the steps NEW ProgressiveBreak into steps, validate each Correct earlyCtrl+C if wrong direction AdaptiveLet the model decide on thinking

Advanced Patterns

Agent View	`claude agents` / `←←` NEW
Background exec	`claude --bg --exec "pytest"` NEW
Worktrees	`claude -w feat --tmux`
Fallback chain	`--fallback-model sonnet,haiku` NEW
Auto Mode	Pro+: classifier handles permissions NEW
From PR	`--from-pr 42`
Schedule	`/schedule` / Routines (web)

Headless / CI

claude -p --max-budget-usd 2 \
  --max-turns 5 \
  --fallback-model opus,sonnet \
  --effort xhigh "fix failing test"

Debug: claude --safe-mode to isolate a faulty customization. NEW

Learning loop: Claude makes a mistake → you correct it → promote it to a MUST rule → Claude never repeats it. That's compound learning.

📝 Customize Claude Code — Build Your Extensions

📋 Slash command — `.claude/commands/name.md`

---
description: Review staged changes
allowed-tools: Read, Bash(git *)
disallowed-tools: Bash(rm *)   # NEW
model: sonnet
argument-hint: "[focus area]"
---
Diff: !`git diff --staged`
Review: $ARGUMENTS | $1, $2 positional

You invoke it: /name. Personal: ~/.claude/commands/

🧠 Skill — `.claude/skills/name/SKILL.md`

---
name: deploy
description: "Deploy to production (1536 chars max)"
context: fork          # isolated subagent
effort: high           # force effort
disallowed-tools: WebFetch  # NEW
user-invocable: true  # false = Claude only
---
Instructions... + [ref](references/guide.md)

Claude decides when to invoke it. /reload-skills re-scans without restart.

🤖 Subagent — `.claude/agents/name.md`

---
name: security-reviewer
description: "Audit code for vulnerabilities"
model: sonnet
tools: Read, Grep, Glob
memory: project        # user / project / local
isolation: worktree    # separate git worktree
---
You are a security specialist...

Invoke: @"name (agent)" | --agent name | natural language

📐 Rule — `.claude/rules/name.md`

---
paths: ["src/**/*.tsx", "src/**/*.ts"]
---
MUST use TanStack Query for data fetching.
SHOULD prefer named exports over default.

Auto-loaded when Claude touches a file matching the glob.

🔗 Hook — `.claude/settings.json`

"hooks": {
  "PostToolUse": [{
    "matcher": "Write|Edit",
    "if": "Edit(*.ts)",
    "hooks": [{
      "type": "command",
      "args": ["prettier", "--write", "$FILE"]
    }]
  }]
}

Deterministic — always runs (unlike CLAUDE.md ~80%).

🔌 Plugin — `claude plugin init <name>`

# Scaffold a plugin (skills+agents+hooks+MCP)
claude plugin init my-plugin
# Load for the session (folder, .zip, URL)
claude --plugin-dir ./my-plugin
claude --plugin-url https://.../plugin.zip

Essentials

Navigation & Display

Editing & Model

Vim Mode

Quick Prefixes

Confirmation Dialogs

Session & Context

Model & Configuration

Project & Analysis

Extensions & Plugins

Advanced & Remote

System & Account

Starting a Session

Model & Reasoning

Context & Directories

Permissions

Output & Formats

System Prompt

Background & Remote

Plugins & Config

CLI Subcommands

1. CLAUDE.md — System Prompt

2. Rules — Scoped Rules

3. Custom Commands — Shortcuts

4. Skills — Complex Workflows

5. Subagents — Parallel Specialists

6. MCP + Plugins

🖥 Computer Use — in the CLI

👁 Monitor tool

Pre-approving Tools

Events (27)

5 Handler Types

Pipe Input

Background & Agent View

Multi-directory & Worktrees

Built-in Agent Tools

Hierarchy (descending priority)

Environment Variables

Settings.json — Key Fields

Project level (.claude/)

Global level (~/.claude/)

Hierarchy (5 levels)

Rewind Options

Context Management

Availability NEW

Activation

Classifier — 2 Paths

Hard deny NEW

Available Models

Fable 5 — automatic fallback NEW SUSPENDED

Effort Levels

Invoke

CLI

Agent View NEW

Desktop app

Web — claude.ai/code

IDE

Open

In the view

claude agents flags

Background sessions

Dynamic workflows (research preview)

Security guidance plugin NEW

Ultrareview NEW

Start a Project

Plan → Execute Loop

Manage Context

Watchers & Automation

Prompting Techniques

Advanced Patterns

Headless / CI

📋 Slash command — .claude/commands/name.md

🧠 Skill — .claude/skills/name/SKILL.md

🤖 Subagent — .claude/agents/name.md

📐 Rule — .claude/rules/name.md

🔗 Hook — .claude/settings.json

🔌 Plugin — claude plugin init <name>

Basic Commands

Models & Effort

Project level (`.claude/`)

Global level (`~/.claude/`)

`claude agents` flags

📋 Slash command — `.claude/commands/name.md`

🧠 Skill — `.claude/skills/name/SKILL.md`

🤖 Subagent — `.claude/agents/name.md`

📐 Rule — `.claude/rules/name.md`

🔗 Hook — `.claude/settings.json`

🔌 Plugin — `claude plugin init <name>`