/ghostPrevents context storage
Ghost Mode
Hide your entire conversation from Claude's training data. Perfect for sensitive projects, confidential work, or testing.
/council1 prompt = multiple views
Council Mode
Get responses from multiple Claude perspectives simultaneously. Great for diverse viewpoints on complex problems.
/researchNo need to paste articles
Research Mode
Enable web search and real-time information gathering. Automatically searches and includes current data in responses.
/focusSmaller context = fewer tokens
Focus Mode
Limit context window to only recent messages. Reduces clutter and forces Claude to be concise in long conversations.
/briefResponse cut 30–50%
Brief Mode
Claude gives shorter, punchier responses automatically. No fluff, no explanations, just output. Perfect for lists and quick answers.
/codeSkips generic explanations
Code Mode
Optimized for coding tasks. Claude focuses on code quality, efficiency, and minimal explanation. Syntax highlighting and execution-ready output.
/creativeFull creative output
Creative Mode
Unlock longer-form, more imaginative responses. For writing, brainstorming, storytelling. Claude is less constrained by brevity.
/deepMore tokens, higher quality
Deep Analysis Mode
Enable extended thinking and complex reasoning. Claude takes more time to analyze before responding. Better for research and strategy.
/nociteRemoves 10–20% overhead
No Citations Mode
Skip source citations in web search results. Saves tokens on citation formatting. Use when sources are less important than content.
/format:jsonEliminates reformatting
Specify Output Format
Directly specify format: /format:json, /format:markdown, /format:csv. Claude won't waste tokens reformatting later.
/length:short30–50% token reduction
Set Response Length
Use /length:short, /length:medium, /length:long. Built-in constraint prevents verbose responses upfront.
/audience:techPrevents re-explanation
Target Audience Setting
Set audience: /audience:executive, /audience:technical, /audience:beginner. Claude auto-adjusts complexity and tone.
/cacheCached = 90% reduction
Enable Response Caching
Cache repeated responses locally. If you ask similar questions, Claude pulls from cache instead of regenerating.
/batch:1040–60% less than separate prompts
Batch Mode
Process multiple items in one response. /batch:10 tells Claude to handle 10 items efficiently. Better than asking separately.
/summarySaves time vs rereading
Auto-Summary Mode
Claude automatically summarizes its own response at the end. Useful for long outputs where you want both detail and overview.
/template20% fewer tokens
Use Template Mode
Apply saved response templates (JSON, markdown, reports). Claude fills in the template structure instead of generating from scratch.
/noexplainRemoves 40–60% of response
Skip Explanations
Tells Claude not to explain its reasoning. Gives output only. Massive token saver for experienced users who understand concepts.
/streamFeels faster
Streaming Mode
Get response token-by-token in real-time instead of waiting for complete response. Better UX, same token count but perceived faster.
/compare:A,B,CTables use 30% fewer tokens
Structured Comparison
Create a comparison table of A vs B vs C automatically. More efficient than asking descriptively.
/list:bulletsNo follow-up reformatting
Optimize List Output
/list:bullets, /list:numbers, /list:prose. Pre-specify format to avoid reformatting requests.
/thinkAdds 20–30% tokens
Show Thinking Process
Claude shows its reasoning step-by-step. Useful for learning or debugging but adds token overhead.
/quiet50–70% shorter responses
Quiet Mode
Ultra-minimal responses. Just the answer, nothing else. Like /brief but even shorter.
/context:5Exponential token savings
Limit Context Window Size
/context:5 limits Claude to last 5 messages only. Forces conciseness in long conversations.
/proFaster, more context
Pro Mode (Claude Pro)
Access Claude's advanced features: longer context, faster processing, priority processing. Requires Claude Pro subscription.
/betaVaries by feature
Access Beta Features
Opt-in to beta/experimental features. New commands or modes available here before public release.
/clearNext prompt doesn't count previous
Clear Conversation History
Start fresh in same thread without creating new conversation. Resets context but keeps you in same chat.
/revertZero new tokens used
Revert to Previous Response
Go back to Claude's previous response without regenerating. Saves tokens vs hitting regenerate.
/edit:textOnly edit tokens, not full
Edit Last Response In-Place
Edit Claude's last response directly instead of asking for a new one. Lighter than regeneration.
/branchNo extra tokens for organisation
Create Conversation Branch
Split conversation at any point to explore alternative responses without losing main thread.
/export:mdNo generation tokens
Export Conversation
/export:pdf, /export:txt, /export:json. Downloads conversation without regenerating content.
/model:haikuHaiku is 90% cheaper
Switch Claude Model
/model:opus, /model:sonnet, /model:haiku. Choose different Claude versions for cost/speed optimization.
/quality:draftDraft uses 40–60% fewer tokens
Set Response Quality
/quality:draft, /quality:standard, /quality:premium. Draft is faster and cheaper.
/custom:formalShorthand vs long instructions
Custom Behavior Rules
Apply custom instruction sets without full system prompts: /custom:no-markdown, /custom:formal.
/memory30–50% savings on follow-ups
Enable Memory
Claude remembers previous conversations with you. You don't need to re-explain context every session.
/shareNo regeneration tokens
Share Conversation Link
Generate shareable link to conversation without regenerating. Read-only access for others.
/version:2024Avoids web search when not needed
Use Specific Knowledge Cutoff
Access Claude's knowledge from a specific date if available. Useful for historical consistency.
/timeout:10Forces concise responses
Set Response Timeout
Force response within X seconds. Claude prioritizes speed over depth if timeout is short.
/confidenceMinimal overhead, useful
Show Confidence Levels
Claude indicates confidence in answers (high/medium/low). Adds metadata without much overhead.
/cite:apaOne-format request
Citation Format Control
/cite:apa, /cite:chicago, /cite:mla. Specifies citation style upfront instead of reformatting.
/voice:casualShorthand avoids re-explaining
Voice/Tone Control
/voice:professional, /voice:casual, /voice:humorous. Sets tone without lengthy explanations.
/privateNo privacy overhead
Private Mode
Conversation data not saved to your account history. Lighter privacy than /ghost.
/limit:500Forces maximum concision
Set Absolute Token Limit
/limit:500 means Claude's response cannot exceed 500 tokens. Hard cap.
/follow-up:offRemoves 2–5% overhead
Disable Follow-Up Suggestions
Claude won't suggest follow-up questions. Saves tokens on UI suggestions.
/examples:off20–30% reduction
Skip Example Output
Claude won't include examples in responses. Saves tokens for users who understand concepts.
/math:latexOne-format math
Math Notation Control
/math:latex, /math:unicode, /math:text. Specifies notation upfront to avoid reformatting.
/coding:pythonAvoids language-switching
Lock Coding Language
/coding:python, /coding:javascript. Claude stays in language instead of asking or mixing.
/api-readyAvoids manual parsing
API-Ready Output Format
Format output for direct API integration (JSON schema, no prose). Perfect for integrations.
/debugHelps you measure
Debug Mode
Claude shows approximate token count of request and response. Useful for optimization.
/archivePrevents context bloat
Auto-Archive Old Conversations
Old conversations automatically archived from active list. Keeps context window fresh.
/smart-cache70–90% savings on repeats
Intelligent Caching (Latest)
Claude automatically caches common patterns. Detects repetition and serves cached responses. Game-changer.