The Complete Claude Code Commands Reference and Pro Tips for Business Users
Every command you should know, plus the tips that separate casual users from power users
Claude Code has dozens of built-in commands that most people never discover. Knowing even a handful of them will make you dramatically more effective — and combining them with the right habits will change how you work with AI entirely.
This guide covers every command worth knowing, organized by what they help you do, plus the pro tips that will save you hours every week.
Claude Code vs Claude Chat: When to Use Which
Before diving into commands, a quick distinction:
- Claude Chat (the web app at claude.ai) is for conversations — asking questions, drafting content, brainstorming. No commands, no files, no code execution.
- Claude Code (the terminal tool or desktop app) is for getting work done — it can read and write files, run commands, manage projects, and execute multi-step tasks. All the commands in this guide live here.
If you are just asking questions or having a conversation, Claude Chat is fine. If you are working on anything that involves files, projects, or multi-step tasks, Claude Code is where you want to be.
Complete Command Reference
Session Management
These commands help you manage your conversation and keep Claude focused during long sessions.
/compact — Compress conversation history to free up context space
Why: Claude has a finite memory window. Long sessions fill it up and Claude starts losing earlier details or gets cut off entirely.
When: Proactively during long working sessions, before you hit limits. Also useful when you notice Claude "forgetting" things you discussed earlier. You can pass custom instructions to control what gets preserved.
Example: /compact preserve the client requirements, pricing details, and the three options we outlined
/recap — Get a summary of the current session
Why: Sessions can get long and complex. You need a way to re-orient without scrolling through hundreds of messages.
When: When you return to a session after a break, after a compaction to verify Claude retained the right context, or when you have gone down a rabbit hole and want to step back to see the big picture.
Example: /recap
/goal — Set a persistent objective for the session
Why: During compactions, Claude can lose sight of the overarching mission and start treating each message in isolation. A goal acts as an anchor that survives compression.
When: At the start of any multi-step task that will require many exchanges. Especially important for sessions you expect to run long enough to need compaction.
Example: /goal Create a competitive analysis report comparing our top 3 competitors across pricing, features, and market positioning
/plan — Enter plan mode to design an approach before executing
Why: Jumping straight into work on complex tasks leads to rework. Plan mode forces alignment on approach before Claude starts making changes.
When: Before any non-trivial work where there are multiple valid approaches, when a task touches many areas, or when you want to see the full scope before committing. See the Pro Tips section below for why you should use this more than you think.
Example: /plan create a quarterly business review presentation covering revenue, churn, and growth metrics
/rename — Rename the current session
Why: Default session names are auto-generated and forgettable. When you have 20+ sessions, you need to find the right one fast.
When: At the start of any session you will want to return to. Best used with /color to create a visual and textual identification system.
Example: /rename q3-competitive-analysis
/clear — Clear the conversation context entirely
Why: Sometimes you want a completely fresh start without the baggage of a long conversation. Different from /compact, which preserves a summary.
When: When switching to a completely unrelated task in the same window, or when the context has become too muddled to salvage.
Example: /clear
/cost — Show token usage and cost for the current session
Why: Usage costs money. You should know what you are spending, especially during long or complex sessions.
When: Periodically during expensive sessions, or at the end of a session to understand what it cost. Pairs well with the model selection strategies in our model guide.
Example: /cost
/help — Built-in help reference
Why: Quick access to documentation without leaving Claude Code.
When: When you forget a command name or want to see what is available.
Example: /help
Planning and Strategy
These commands help you think before you act — essential for complex work.
/goal and /plan are covered above in Session Management since they serve both purposes.
/diff — Show what Claude has changed in your files
Why: Quick way to see every edit Claude made without scrolling through the full conversation.
When: Before you finalize any work. Think of it as a sanity check — review everything Claude touched before you accept it.
Example: /diff
Review and Quality Commands
These commands help you catch mistakes before they go live. If you work with a developer or technical team, these are especially relevant — but they also apply to any project where Claude is editing files.
/code-review — Have Claude review changes for issues
Why: A second set of eyes catches mistakes. This is Claude reviewing its own work (or your team's work) before you finalize anything.
When: Before accepting any significant changes Claude has made. Think of it as asking a colleague to double-check the work.
Example: /code-review
/simplify — Clean up and streamline what Claude just built
Why: First drafts — whether code, documents, or workflows — often have redundancy. This trims the fat.
When: After Claude finishes a task, before you finalize. Focuses on making things cleaner and more efficient.
Example: /simplify
/review — Review a pull request
Why: If your team uses GitHub, Claude can analyze proposed changes and give you a summary of what is being changed and whether it looks right — even if you are not technical.
When: When your developer submits changes and you want to understand what they did before approving.
Example: /review
/security-review — Check for security vulnerabilities
Why: Security issues are easy to miss. This specifically looks for common vulnerabilities like exposed data, weak authentication, or unsafe handling of user input.
When: Before launching any changes to your website, app, or customer-facing tools. Ask your developer to run this as a standard practice.
Example: /security-review
/init — Create a project memory file (CLAUDE.md)
Why: CLAUDE.md is a file that tells Claude about your project — what it does, how it is structured, your preferences. Without it, Claude starts from scratch every session. With it, Claude already knows the context.
When: Once per project to set it up. Then maintain it as your project evolves. Think of it as onboarding documentation for Claude.
Example: /init
/run — Launch your project and see it in action
Why: Changes that look correct on paper can still behave unexpectedly. This runs your project so you (or Claude) can observe what actually happens.
When: After making changes and you want to see the real result — not just trust that it works.
Example: /run
/verify — Confirm a change actually works as expected
Why: "Done" and "working correctly" are not the same thing. Verify tests the change and checks that the result matches what you asked for.
When: After any change where you want proof it works — especially for anything customer-facing.
Example: /verify
Configuration and Setup
These commands control how Claude Code looks and behaves.
/model — Switch between Claude models mid-conversation
Why: Different tasks have different cost, speed, and quality tradeoffs. You should not pay premium prices for simple tasks. See our full model guide for when to use each.
When: Drop to Haiku for simple tasks, use Sonnet for everyday work, step up to Opus for complex analysis.
Example: /model haiku then later /model opus
/color — Change the accent color of your prompt bar
Why: Humans are visual. Color-coding helps you stay organized across multiple sessions.
When: Create a system. Green for data analysis, orange for general questions, red for production issues — build your own. Best used with /rename so sessions have both visual and textual identity.
Example: /color red
/statusline — Configure what info appears in the status bar
Why: Key metrics like token usage, cost, and current model are useful to monitor passively without having to ask.
When: Run once to set your preferred layout. Revisit if you change workflows.
Example: /statusline
/terminal-setup — Configure shell integration for Claude Code
Why: Features like the status line and notifications depend on your terminal being set up correctly.
When: Once, when you first install Claude Code in a new terminal environment. Re-run if you switch terminals.
Example: /terminal-setup
/config — Open or modify Claude Code settings
Why: Claude Code's behavior, permissions, and preferences live in a settings file. This is the direct way to manage them.
When: When you want to adjust what Claude is allowed to do, change default behaviors, or tweak settings without editing files manually.
Example: /config
/fast — Toggle fast mode for quicker responses
Why: Sometimes you want speed over thoroughness. Fast mode gives you faster responses without downgrading the model.
When: When you are going back and forth quickly on small changes and want Claude to respond faster. Toggle off when you need Claude to think harder on a complex task.
Example: /fast
Advanced Commands
These commands extend what Claude can do and automate your workflows.
/hooks — Set up automated actions that trigger on Claude events
Why: If you find yourself telling Claude "and also do X" after every task, hooks automate that. They are rules that run automatically — like "every time Claude edits a file, also check it for errors."
When: When you notice a repetitive pattern in your sessions. Hooks turn a manual reminder into an automatic guardrail.
Example: /hooks
/mcp — Manage your plugin and tool connections
Why: This is how you install and manage the integrations that connect Claude to Slack, GitHub, Figma, and hundreds of other tools. See our guides on plugins and connectors for the full picture.
When: When setting up a new tool integration, when a connection drops and you need to reconnect, or when you want to see what is currently connected.
Example: /mcp
/skill — Run a specialized workflow by name
Why: Skills are pre-built workflows for common tasks — creating PDFs, generating presentations, running security audits, building reports. Instead of explaining what you want step by step, one command triggers the whole workflow.
When: When you need to do something that sounds like a common task. Check what skills are available — you might save yourself ten minutes of prompting.
Example: /skill pdf to create a PDF, or /skill pptx to build a presentation
/agents — View or manage available sub-agents
Why: For large tasks, Claude can spin up specialized helpers that work on different parts of the problem simultaneously — like having a team instead of one person. Sub-agents keep the main conversation clean while work happens in the background.
When: When you want to see what types of specialized helpers are available, or when you are curious how Claude is breaking down a complex task behind the scenes.
Example: /agents
/loop — Run a task on a recurring interval
Why: Some tasks need repeated checking — waiting for a report to finish, monitoring a process, tracking a metric. Loop tells Claude to keep checking so you do not have to.
When: When you need something monitored or checked on a schedule. Omit the interval to let Claude decide how often to check.
Example: /loop 10m check the website and let me know if it is back online
/schedule — Create and manage scheduled remote agents
Why: Some tasks need to run on a schedule without you being present — daily reports, periodic checks, recurring maintenance.
When: When you want Claude to do something on a cron schedule that runs automatically.
Example: /schedule create "every Monday at 8am, pull last week's sales data and draft a weekly summary report"
Pro Tips That Will Save You Hours
For Large Inputs: Use File-Based Workflows
When you have a lot of text to give Claude — a long document, a data dump, meeting transcripts — do not paste it directly into the chat. Instead:
- Save the content to a file
- Ask Claude to read the file
This is more reliable, avoids truncation issues, and lets Claude reference the file multiple times without you re-pasting.
Bonus tip: When you need complex visual output (reports, dashboards, comparisons), ask Claude to generate HTML instead of markdown. The results are dramatically better. Anthropic wrote about this approach in their blog post on the unreasonable effectiveness of HTML in Claude Code.
For Better Prompts: Let Claude Help You Write Them
Here is a secret most people miss: Claude is better at writing prompts than you are. Instead of struggling to phrase your request perfectly, tell Claude what you want to achieve and ask it to help you construct the prompt.
You can even ask it to use specific prompting patterns. For example, the persona pattern:
"I need a prompt that will help me evaluate job candidates for a product manager role. Use a persona pattern — something like 'Act as a VP of Product with 15 years of experience, evaluate this candidate based on...' Help me construct this."
Claude knows dozens of prompting patterns and will build a better prompt than you would writing it from scratch. Use Claude to prompt Claude.
For Longer Sessions: Always Start with Plan Mode
This is the single most impactful habit you can build. For any task that is not a one-line question, start with /plan.
Here is the workflow:
- Type
/planand explain your goal. Be specific about what you want to accomplish and what "done" looks like. - Ask Claude to ask you clarifying questions. It usually does this automatically, but if there is a specific area you are unsure about, call it out.
- Tell Claude how to verify its work. Define what success looks like — "the tests should pass," "the report should include these three sections," "the output should match this format."
- Have Claude output the plan to a document with clear phases. This becomes your shared roadmap. Reference it throughout the session. Have Claude keep it updated as phases complete.
Think of this like fully delegating a task to a colleague. You would not just say "handle it" — you would align on the goal, clarify the approach, set checkpoints, and define what done looks like. Do the same with Claude.
This approach is powerful because Claude will:
- Surface edge cases you had not considered
- Ask clarifying questions before making assumptions
- Break complex work into manageable phases
- Check in with you at decision points
- Verify its own work against the criteria you defined
For Predictable File Output: Set Up an Output Directory
When working in Claude Code, tell Claude where to put generated files. A clear /output directory (or whatever structure makes sense for your project) means:
- You always know where to find what Claude created
- Files do not end up in random locations
- You can set this expectation once and it carries through the session
Simply tell Claude at the start of a session: "Put all generated files in the /output directory" or set it up as part of your project configuration.
For Structured Results: Tell Claude Your Desired Format
Claude will default to whatever format seems reasonable. But if you tell it exactly what you want, the results are dramatically more useful.
Specify the file format: Markdown, HTML, JSON, CSV, PDF — tell Claude upfront.
Specify the structure within the format. For example, instead of "give me interview questions," try:
"Give me 10 interview questions for a senior product manager. For each question, use this format:
- Question:
- When to ask:
- Why ask this:
- What a good answer sounds like:
- Red flags to watch for:"
The more specific your structure, the more useful the output. This works for any type of content — reports, analyses, action items, research summaries.
The Five-Minute Setup for Every Session
Before diving into work, spend five minutes on this checklist:
/renameyour session to something meaningful/colorit to match your system/goalto set the session objective/planto align on approach before executing/modelto pick the right model for today's work (default to Sonnet)
This small investment compounds across every session. You will stay organized, spend less, and get better results.
Common Mistakes and How to Avoid Them
Mistake: Starting work without a plan.
Fix: Use /plan for anything beyond a simple question. The five minutes you spend planning saves thirty minutes of rework.
Mistake: Never checking your costs.
Fix: Run /cost at the end of every session. You will quickly learn which tasks are expensive and adjust your model selection.
Mistake: Using the same model for everything. Fix: Switch models deliberately. Most people should be on Sonnet 80% of the time with occasional moves to Haiku or Opus.
Mistake: Pasting huge blocks of text directly into chat. Fix: Save to a file and ask Claude to read it. The results are more reliable and you can reference the file multiple times.
Mistake: Not defining what "done" looks like. Fix: Tell Claude how to verify its work. "The report should cover X, Y, and Z," "compare the output to this example," "make sure the final version includes all three client requirements." Verification criteria make everything better.
This article is part of The Complete Guide to Claude AI for Business Owners. For model selection strategies, see our Model Selection Guide.