Claude Code Commands: 100 Worth Knowing and the 20 I Use Daily

I counted. Claude Code now has well over a hundred commands, flags and shortcuts, and most lists online mix in commands that do not exist, like /commit or /pr, or ones Anthropic removed months ago. So I went through Anthropic’s official command, CLI and interactive mode references in October 2026 and put together the 100 that are actually worth knowing.

Then I did the more useful thing: I picked the 20 I use every single day as a QA engineer. If you only learn one part of this post, learn that part. The full list is there for when you need it.

A note before we start. Claude Code ships updates often, and some commands depend on your plan or platform. Type / in a session to see exactly what your version supports, and run /release-notes when something looks different.

The 20 Claude Code commands I use every day

These are ordered roughly by how often my fingers hit them. Each one comes with the QA reason it earned its place.

#	Command	Why it is in my top 20
1	Ctrl+V	Paste a screenshot of the bug straight into the prompt. Claude reads the image, and I can refer to it as [Image #1]. Half my bug reports start this way.
2	@path/to/file	Mention a file so Claude reads exactly the spec, page object or fixture I mean, instead of guessing from the folder name.
3	! npx playwright test	Shell mode. The command runs, its output lands in the conversation, and Claude responds to the real failure, not my summary of it.
4	Shift+Tab	Cycles permission modes. I go to plan mode for anything risky and auto-accept edits for boring refactors of test data.
5	/plan	Plan mode straight from the prompt, for example /plan add API tests for the refunds endpoint. Claude reads and proposes before it edits.
6	/init	Creates a CLAUDE.md for a new repo. The first thing I run on any test framework I inherit.
7	/memory	Edits CLAUDE.md and auto memory. This is where rules like “never use waitForTimeout” live so I stop repeating them.
8	/model	Switches the model and saves it as default. A quick fixture change does not need the biggest model; a nasty race condition does.
9	/context	Shows a colored grid of what is filling the context window. When answers get vague, this usually tells me why.
10	/compact	Summarizes the conversation so far and frees context. I pass focus instructions like /compact keep the failing test names.
11	/clear	Starts a fresh conversation. I clear between unrelated tasks so a checkout bug does not leak into a login test.
12	/rewind	Rolls the conversation or the code back to an earlier point. My undo button when Claude goes down the wrong path.
13	/diff	Shows every change in the working tree, including Claude’s edits. I read this before anything gets committed.
14	/code-review	Reviews the current diff, a branch or a PR for correctness. Effort levels go from low to max, and ultra runs a deep cloud review.
15	/security-review	Checks the changes on my branch for injection, auth issues and data exposure. Cheap insurance before a release.
16	/verify	Builds and runs the app to confirm a change actually works, rather than trusting that the tests passed.
17	claude -c	Continues the most recent conversation in the current folder. My first command most mornings.
18	/resume	Opens the session picker, or resumes one by name. I name sessions with /rename so I can find them later.
19	/loop	Runs a prompt repeatedly, for example watching a CI run or rerunning a flaky spec until I understand the pattern.
20	/mcp	Manages MCP servers and their logins. I check here first when Playwright or Jira tools stop responding.

If you are new to MCP, our explainer on why MCP exists when we already had APIs is a good primer before you connect Playwright or Jira.

---

All 100 Claude Code commands, flags and shortcuts

Grouped by what you are trying to do. Slash commands run inside a session. CLI commands and flags run from your terminal. Shortcuts are keys you press in the prompt.

Session and context management

#	Command	What it does	When I reach for it
1	/clear [name]	Starts a new conversation with empty context; the old one stays resumable	Switching to an unrelated task
2	/compact [instructions]	Summarizes the conversation to free context, with optional focus	Long debugging sessions
3	/context	Visualizes context usage as a grid with suggestions	Answers get vague or slow
4	/autocompact [auto|tokens]	Sets how full context gets before automatic compaction	Tuning very long sessions
5	/resume [session]	Resumes a conversation by ID or name, or opens the picker	Coming back to yesterday’s work
6	/rename [name]	Names the current session	Before a session you will want back
7	/rewind	Rewinds the conversation and/or code to a previous point (aliases /checkpoint, /undo)	Claude took a wrong turn
8	/branch [name]	Branches the conversation so you can try another direction	Comparing two fixes
9	/fork [prompt]	Copies the conversation into a new background session	Running a side investigation
10	/btw [question]	Asks a side question without adding it to the conversation	Quick lookups mid-task
11	/recap	Generates a one-line summary of the session	Status updates for standup
12	/export [filename]	Exports the conversation as plain text	Attaching an investigation to a ticket
13	/copy [N]	Copies the last (or Nth-latest) response, or a single code block	Grabbing a generated test
14	/focus	Shows only your prompt, a tool summary and the final answer	Reducing noise
15	/cd <path>	Moves the session to another directory, keeping the conversation	Jumping between repos
16	/add-dir <path>	Adds a working directory for file access	Tests live in a separate repo

Models, effort and cost

#	Command	What it does	When I reach for it
17	/model [model]	Switches the model and saves it as your default	Matching model to task difficulty
18	/effort [level]	Sets effort from low to max, or auto	Hard bugs versus quick edits
19	/fast [on|off]	Toggles fast mode	When speed matters more than depth
20	/advisor [model|off]	Enables a second model that advises at key moments	Tricky architecture decisions
21	/usage	Shows session cost, plan limits and activity (aliases /cost, /stats)	Checking spend
22	/output-style [style]	Lists or switches output styles, like concise	Shorter answers

Planning, agents and automation

#	Command	What it does	When I reach for it
23	/plan [description]	Enters plan mode, optionally with a task	Anything that touches many files
24	/goal [condition]	Keeps Claude working across turns until a condition is met	Get the suite green, with limits
25	/loop [interval] [prompt]	Runs a prompt repeatedly while the session is open	Watching CI or a flaky test
26	/schedule [description]	Creates or manages routines that run in the cloud	Nightly test health reports
27	/tasks	Views and manages background work (alias /bashes)	Checking long test runs
28	/subtask <task>	Spawns a forked subagent that inherits the conversation	Parallel investigation
29	/background [prompt]	Detaches the session to run as a background agent	Long refactors
30	/batch <instruction>	Splits a large change into parallel units after a plan	Migrating hundreds of tests
31	/workflows	Opens the workflow progress view	Watching multi-step runs
32	/deep-research <question>	Runs web searches and writes a cited report	Comparing test tools
33	/agents	Reminds you how to create and manage subagents	Setting up a test-writer agent
34	/list-agents	Lists subagents and sessions Claude can message	Coordinating parallel work

Code review, testing and verification

#	Command	What it does	When I reach for it
35	/diff	Reviews changes in the working tree, including Claude’s edits	Before every commit
36	/code-review [effort] [target]	Reviews the diff, a PR, branch or path for correctness (alias /review)	Pre-merge checks
37	/security-review	Analyzes branch changes for vulnerabilities	Before releases
38	/ultrareview [PR or branch]	Runs a deep multi-agent review in a cloud sandbox	High-risk PRs
39	/simplify [target]	Reviews changed code for cleanup and applies fixes	Tidying test helpers
40	/verify	Builds and runs your app to confirm a change works	Proving a fix, not just a green test
41	/run	Launches and drives your app to see a change working	Manual-style exploratory checks
42	/run-skill-generator	Teaches /run and /verify how to build your project	First-time setup
43	/debug [description]	Enables debug logging and reads the session log	Claude Code itself misbehaves
44	/autofix-pr [prompt]	A cloud session watches your PR and pushes fixes for CI failures and review comments	Keeping a PR green
45	/claude-in-chrome [task]	Has Claude act in your browser: test a page, fill a form, read console logs	Quick UI checks

Project setup, memory and permissions

#	Command	What it does	When I reach for it
46	/init	Creates a CLAUDE.md guide for the project	New repo
47	/memory	Edits CLAUDE.md files and auto memory	Recording team conventions
48	/permissions	Manages allow, ask and deny rules for tools	Locking down risky commands
49	/fewer-permission-prompts	Builds an allowlist of common read-only commands from your history	Too many prompts
50	/hooks	Views hook configurations	Auto-running lint after edits
51	/config	Opens settings, or sets key=value directly	Theme, model, editor mode
52	/update-config [request]	Describe a settings change and Claude edits settings.json	Adding an allowed command
53	/sandbox	Toggles sandbox mode where supported	Running untrusted scripts
54	/import [source]	Brings config from Codex, Gemini CLI or Cursor	Switching tools
55	/status	Shows version, model, account and connectivity	Support questions
56	/doctor	Runs a setup checkup that can fix issues	Something is broken
57	/login	Signs in to your Anthropic account	First run

Skills, plugins, MCP and integrations

#	Command	What it does	When I reach for it
58	/skills	Lists available skills, sortable by token cost	Seeing what is installed
59	/reload-skills	Rescans skill folders without restarting	After editing a skill
60	/skill-doctor	Shows each skill’s context cost and usage	Trimming unused skills
61	/plugin [subcommand]	Manages plugins: list, install, enable, disable	Adding team plugins
62	/reload-plugins	Reloads plugins to apply changes	After plugin edits
63	/mcp	Manages MCP servers and OAuth logins	Tool connection problems
64	/ide	Manages IDE integrations	VS Code or JetBrains setup
65	/chrome	Configures Claude in Chrome	Browser testing setup
66	/install-github-app	Installs the Claude GitHub App and optional Actions	PR automation
67	/claude-api	Loads Claude API reference and workflows like migrate and cost-optimize	Building on the API

Interface, help and housekeeping

#	Command	What it does	When I reach for it
68	/help	Shows help and available commands	Always
69	/release-notes	Shows the changelog by version	After an update
70	/insights	Generates an HTML report on how you use Claude Code	Monthly self-review
71	/team-onboarding	Generates an onboarding guide from your usage history	New teammate joins
72	/keybindings	Opens your keyboard shortcuts file	Custom keys
73	/statusline	Configures the status line	Showing branch and model
74	/theme	Changes the color theme, including colorblind themes	Readability
75	/terminal-setup	Installs Shift+Enter for newlines in supported terminals	Multiline prompts
76	/teleport	Pulls a cloud session into your terminal (alias /tp)	Continuing web work locally
77	/remote-control	Makes the session controllable from claude.ai	Checking in from your phone
78	/desktop	Continues the session in the Desktop app (alias /app)	Prefer a GUI
79	/feedback	Sends product feedback to Anthropic	Reporting an issue

Terminal (CLI) commands and flags

#	Command	What it does	When I reach for it
80	claude	Starts an interactive session	Every day
81	claude "query"	Starts a session with an initial prompt	Jumping straight in
82	claude -p "query"	Runs one query non-interactively and exits	Scripts and CI
83	cat log.txt | claude -p "explain"	Pipes content in	Explaining a failure log
84	claude -c	Continues the most recent conversation here	Picking up where you left off
85	claude -r "name" "query"	Resumes a session by ID or name	Returning to a named investigation
86	claude -w name	Starts in an isolated git worktree	Parallel work on one repo
87	claude --bg "task"	Starts a background agent and returns immediately	Fire and forget investigations
88	claude agents	Opens agent view to monitor background sessions	Managing parallel agents
89	claude --permission-mode plan	Starts in a chosen permission mode	Read-only exploration
90	claude --model sonnet	Sets the model for this session	Cheaper scripted runs
91	claude -p --output-format json "query"	Prints JSON output for scripts	Parsing results in CI
92	claude -p --max-turns 3 "query"	Limits agentic turns in print mode	Bounded CI jobs
93	claude update	Updates to the latest version	Weekly
94	claude doctor	Prints installation diagnostics without starting a session	Install problems
95	claude mcp	Configures MCP servers from the terminal	Scripting tool setup

Keyboard shortcuts and prompt prefixes

#	Command	What it does	When I reach for it
96	Shift+Tab	Cycles permission modes, including plan and auto-accept	Changing how much Claude can do
97	Esc	Interrupts Claude mid-response, keeping work done so far	Redirecting quickly
98	Esc Esc	Clears your draft, or opens rewind on an empty prompt	Undoing
99	! (at start)	Shell mode: runs a command and adds its output	Running tests inline
100	@ (in prompt)	File path mention with autocomplete	Pointing at exact files

That is 100. If you want to see how these fit into a real testing workflow, our post on context engineering for AI testing explains why /context, /memory and @ mentions matter so much for test quality.

---

Commands that no longer exist

Many cheat sheets still list these. Save yourself the confusion:

Command	Status	Use instead
/vim	Removed in v2.1.92	/config, then Editor mode
/pr-comments	Removed in v2.1.91	Ask Claude to read the PR comments
/ultraplan	Removed	Plan mode with /plan or Shift+Tab
/commit, /pr	Never built in	Ask Claude to commit or open a PR, or write your own skill

How to make your own commands

Built-in commands only go so far. Any repeated QA task, like “write API tests for this endpoint using our conventions”, can become your own command. In Claude Code these are skills: a folder with a SKILL.md file under .claude/skills/ in your repo, or ~/.claude/skills/ for yourself. Once it exists, you call it with /its-name, and /reload-skills picks up changes without a restart.

Three skills worth building for a QA team:

• /api-tests: generates API tests for a route following your folder layout, naming and assertion style.
• /bug-report: turns a failing test, screenshot and log into a ticket with steps to reproduce.
• /flaky-check: runs a spec several times, compares failures and labels the likely cause.

If you want a structured path from here, our Anthropic MCP course path for test engineers picks up where this post ends.

Start with five

Do not try to learn 100 commands. Learn these five this week: Shift+Tab, @, !, /compact and /diff. They change how you work more than the other 95 combined. Add one new command a week after that, and in a couple of months you will have your own top 20.

Tell me yours in the QABash community. I am always curious which commands other testers cannot live without.

Bookmark