CLI for Better prompts, transparent costs, & zero vendor lock-in. Optimize AI prompts across OpenAI, Claude, Gemini, Grok, xAI, and DeepSeek with detailed per-token pricing. BYOK keeps it honest.

Star us on GitHub | Request a Feature

🔑 BYOK (Bring Your Own Key) • 🎨 16 Beautiful Themes • ⚡ Lightning Fast

📦 Installation

Get started in seconds:

npm install -g megabuff
megabuff optimize "Rewrite this prompt to be clearer"

# Use interactive mode for multiple commands!
megabuff shell

# Set up your API key (choose your favorite provider)
megabuff config token YOUR_API_KEY --provider openai

That's it! 🎉 You're ready to supercharge your prompts.

✨ Features

• 🤖 Multi-Provider Support - OpenAI, Anthropic Claude, Google Gemini, xAI, DeepSeek & more
• 🔑 BYOK Model - Bring your own API key, full control
• 🐚 Interactive Shell Mode - Run multiple commands without repeating megabuff (NEW!)
• 🎨 16 Beautiful Themes - Customize your CLI experience
• 🎭 7 Optimization Styles - Concise, detailed, technical, creative, formal, casual, balanced
• 🔧 Custom System Prompts - Ultimate control over optimization behavior
• 🔄 Iterative Refinement - Progressive improvement with multiple optimization passes
• 🔍 Comparison Mode - Test multiple providers side-by-side to find the best result
• 📊 Prompt Analysis - Get detailed feedback on strengths, weaknesses, and improvement suggestions
• 💰 Cost Tracking - Estimate and monitor API costs before running operations
• ⚡ Lightning Fast - Optimize prompts in seconds
• 📋 Auto-Clipboard - Results copied automatically
• 🔀 Flexible Input - Inline, file, pipe, or interactive
• 💾 Multiple Output Formats - Stdout, file, or interactive view
• 🔒 Secure Storage - Keychain support for API keys
• 🎯 Smart Model Selection - Auto-detects provider from model name
• 📊 Stats Tracking - See word count changes and improvements
• 📝 Error Logging - All errors saved to ~/.megabuff/errors.log for easy debugging
• 🌈 Beautiful Output - Themed, formatted, fun to use!

📚 Table of Contents

• ✨ Features
• 📦 Installation
• ⚡ Quick Start
• 🔐 Setup
  • 🤖 Getting Your OpenAI API Key (BYOK)
  • 🧠 Getting Your Anthropic API Key (BYOK)
  • ✨ Getting Your Google Gemini API Key (BYOK)
  • 🚀 Getting Your xAI API Key (BYOK)
  • 🔮 Getting Your DeepSeek API Key (BYOK)
  • Configuring Your API Key
• ⚙️ Configuration Commands
• 🎨 Theme Commands
• 💡 Usage
• 🎯 Examples
• 🔧 How It Works
• 🆚 VS Code Integration
• 🤝 Contributing
• 📄 License
• 👨‍💻 Development

⚡ Quick Start

New to MegaBuff? Let's get you optimizing in under 60 seconds! ⏱️

# 1. Install globally
npm install -g megabuff

# 2. Set up your API key (choose your favorite provider)
megabuff config token YOUR_API_KEY --provider openai

# 3. Start optimizing!
megabuff optimize "Write a function to validate emails"

# 4. Or use interactive mode for multiple commands!
megabuff shell

💡 Pro Tip: Use megabuff shell for an interactive session with guided wizards!

🔐 Setup

🤖 Getting Your OpenAI API Key (BYOK)

MegaBuff uses a BYOK (Bring Your Own Key) model. Why is this awesome?

• ✅ Full Control - You manage your usage and costs
• ✅ Super Cheap - Typically just pennies per optimization 💰
• ✅ Privacy First - Your prompts go directly to the provider 🔒
• ✅ Your Rules - Set your own usage limits

Get your key in 4 easy steps:

1. Create an OpenAI Account

   • Sign up or log in at platform.openai.com
   • Note: This is separate from the standard ChatGPT consumer site

2. Set up Billing

   • The API runs on a pay-per-use model
   • Add a payment method in the Billing section
   • You can set usage limits to manage costs
   • Typical cost: ~$0.001-0.01 per prompt optimization (using gpt-4o-mini)

3. Generate Your API Key

   • Navigate to API Keys in the sidebar
   • Click "+ Create new secret key"
   • Give your key a descriptive name (e.g., "MegaBuff CLI")
   • Click "Create secret key"

4. Save Your Key Immediately

   • Important: Copy the key right away - OpenAI only shows it once!
   • Store it securely - you'll need to generate a new one if you lose it
   • The key starts with sk-

🧠 Getting Your Anthropic API Key (BYOK)

Want to use Claude? Follow these simple steps:

1. Create an Anthropic Console account

   • Sign up / log in at https://console.anthropic.com/

2. Set up billing / credits

   • Ensure your Anthropic account is enabled for API usage (billing/credits as required by Anthropic).

3. Create an API key

   • In the Anthropic Dashboard Console, go to API Keys
   • Click Create key
   • Name it something like "MegaBuff CLI"

4. Save your key immediately

   • Copy and store it somewhere secure
   • Anthropic keys typically start with sk-ant-

✨ Getting Your Google Gemini API Key (BYOK)

Ready to try Gemini? Here's how:

1. Go to Google AI Studio

   • Visit Google AI Studio
   • Sign in with your Google account

2. Create an API key

   • Click Get API Key or Create API Key
   • Choose an existing Google Cloud project or create a new one
   • Your API key will be generated

3. Save your key immediately

   • Copy and store it securely
   • You'll need this key to use Gemini models

4. Enable billing (if needed)

   • Free tier is available for testing
   • For production usage, you may need to enable billing in Google Cloud Console

🚀 Getting Your xAI API Key (BYOK)

Want to use Grok? Here's how to get started:

1. Create an xAI account

   • Visit x.ai or the xAI Console
   • Sign up or log in with your account

2. Access the API Console

   • Navigate to the API section in the xAI Console
   • You may need to join the API waitlist or request access

3. Generate your API key

   • Click Create API Key or Generate Key
   • Give your key a descriptive name (e.g., "MegaBuff CLI")
   • Copy the key immediately

4. Save your key immediately

   • xAI keys typically start with xai-
   • Store it securely - you won't be able to view it again

🔮 Getting Your DeepSeek API Key (BYOK)

Ready to use DeepSeek? Follow these steps:

1. Create a DeepSeek account

   • Visit DeepSeek Platform
   • Sign up or log in to your account

2. Set up billing

   • Navigate to the billing section
   • Add payment method or credits as required
   • DeepSeek offers competitive pricing for API usage

3. Generate your API key

   • Go to API Keys section in the dashboard
   • Click Create new key or Generate API key
   • Name it something like "MegaBuff CLI"

4. Save your key immediately

   • DeepSeek keys typically start with sk-
   • Copy and store it securely
   • You won't be able to view the full key again after creation

Configuring Your API Key

Once you have your provider API key, configure it using one of these methods:

Option 1: Interactive Setup (Recommended)

The easiest way to get started:

# Interactive configuration menu
megabuff config

# Then choose:
# 1) Set API token for a provider
# Pick OpenAI, Anthropic, or Google, paste your key, choose storage method

# Or use the direct token command
megabuff config token sk-your-api-key-here --provider openai
megabuff config token sk-ant-your-key --provider anthropic --keychain
megabuff config token your-google-key --provider google

This saves your key for future use. You only need to do this once!

Storage Options:

• Config file (default): ~/.megabuff/config.json
• System keychain (more secure): macOS Keychain, Windows Credential Manager, or Linux Secret Service

Option 2: Environment Variable

export OPENAI_API_KEY="sk-your-api-key-here"
export ANTHROPIC_API_KEY="sk-ant-your-api-key-here"
export GOOGLE_API_KEY="your-google-api-key-here"
export XAI_API_KEY="xai-your-api-key-here"
export DEEPSEEK_API_KEY="sk-your-deepseek-key-here"

Add to your shell profile (.bashrc, .zshrc, etc.) to persist across sessions.

Option 3: Pass as Flag

megabuff optimize "your prompt" --api-key sk-your-key-here
megabuff optimize --provider anthropic "your prompt" --api-key sk-ant-your-key-here
megabuff optimize --provider google "your prompt" --api-key your-google-key-here

API Key Priority

The CLI checks for your token in this order (per provider):

1. --api-key flag (highest priority)
2. Provider env var (e.g. OPENAI_API_KEY, ANTHROPIC_API_KEY, XAI_API_KEY, DEEPSEEK_API_KEY)
3. System keychain (if configured)
4. Config file at ~/.megabuff/config.json

⚙️ Configuration Commands

🎮 Interactive Config Menu

The easiest way to configure everything:

megabuff config

# Shows:
# ╭─────────────────────────────────────╮
# │   MegaBuff Configuration Setup     │
# ╰─────────────────────────────────────╯
#
# What would you like to configure?
#
#   1) Set API token for a provider
#   2) Set default provider
#   3) Set model (auto-selects provider)
#   4) View current configuration
#   5) Exit

🔧 Direct Commands

Token Management:

# 🔑 Set your API token
megabuff config token sk-your-api-key --provider openai
megabuff config token sk-ant-... --provider anthropic --keychain  # More secure!

# 🗑️ Remove a token
megabuff config remove --provider anthropic

Provider & Model Selection:

# 🤖 Set default provider
megabuff config provider anthropic
megabuff config provider                  # Show current

# 🎯 Set model (auto-selects provider!)
megabuff config model claude-sonnet-4-5   # → Anthropic
megabuff config model gpt-4o              # → OpenAI
megabuff config model gemini-1.5-pro      # → Google
megabuff config model                     # Show current

View Everything:

# 📊 Show complete configuration
megabuff config show

🤖 Available Models

OpenAI:

• gpt-5.2 (default)
• gpt-4o
• gpt-4o-mini
• gpt-4-turbo
• gpt-4
• gpt-3.5-turbo

Anthropic:

• claude-opus-4-5 (default)
• claude-sonnet-4-5
• claude-sonnet-4-0
• claude-sonnet-4-5-20250929
• claude-3-5-sonnet-20241022
• claude-3-opus-20240229
• claude-3-sonnet-20240229
• claude-3-haiku-20240307

Google Gemini:

• gemini-2.5-flash (default)
• gemini-2.5-pro
• gemini-2.5-flash-lite
• gemini-2.0-flash
• gemini-2.0-flash-lite
• gemini-3-flash-preview
• gemini-3-pro-preview
• gemini-flash-latest

xAI (Grok):

• grok-beta (default)
• grok-vision-beta

DeepSeek:

• deepseek-chat (default)
• deepseek-reasoner

🎨 Theme Commands

Make your CLI experience uniquely yours! Choose from 16 stunning color themes. ✨

👀 View Current Theme

megabuff theme

Shows your active theme with a beautiful preview! 🌈

📋 List All Themes

megabuff theme list

Browse all 16 themes with:

• ⭐ Your active theme highlighted
• Live color previews
• Theme descriptions

🎭 Change Theme

megabuff theme set dracula         # 🧛‍♂️ Dark & mysterious
megabuff theme set cyberpunk       # 🌆 Neon future vibes
megabuff theme set pastel-rainbow  # 🌈 Soft & dreamy

Your choice is saved and applied to all commands instantly!

🔍 Preview Theme

Try before you apply:

megabuff theme preview monokai

See the full color palette in action before committing! 🎨

🌟 Available Themes

| Theme | Vibe | Description | |-------|------|-------------| | 🎯 default | Clean & Pro | Cyan and green palette | | 🎨 material | Modern | Google Material Design | | 🧛‍♂️ dracula | Dark & Gothic | Purple and pink accents | | ❄️ nord | Arctic | Cool bluish tones | | ☀️ solarized | Precision | Perfect balance | | 🎸 monokai | Vibrant | Monokai Pro inspired | | 🌆 cyberpunk | Futuristic | Neon city lights | | 🌅 sunset | Warm | Oranges & purples | | 🌈 pastel-rainbow | Dreamy | Soft rainbow hues | | 🍬 bubblegum | Sweet | All the pink vibes | | 🍭 cotton-candy | Fluffy | Pink meets blue | | 🦄 unicorn | Magical | Pastel paradise | | 🌊 ocean | Aquatic | Deep blue serenity | | 🌲 forest | Natural | Earthy greens | | 📺 retro | Nostalgic | 80s terminal style | | ⚡ neon-dreams | Electric | Vibrant neons |

Every theme is optimized for readability and style! 🎨✨

💡 Usage

MegaBuff is flexible! Use whichever input method works best for you:

🐚 Interactive Shell Mode (NEW!)

Run multiple commands without typing megabuff every time!

megabuff shell
# or: megabuff interactive
# or: megabuff i

# Now you're in the shell:
megabuff> optimize "Write a REST API" --style technical
megabuff> analyze "Create a chatbot"
megabuff> theme set cyberpunk
megabuff> exit

Features:

• ⚡ Faster workflow - Skip typing megabuff for every command
• 🧙 Guided wizards - Type optimize or analyze without arguments to start a step-by-step wizard
• 📜 Command history - Use arrow keys (↑/↓) to recall previous commands
• 🔄 Session persistence - API keys and config loaded once
• 📝 Error logging - All errors saved to ~/.megabuff/errors.log for debugging
• 💡 Built-in help - Type help anytime
• 🧹 Clear screen - Type clear to refresh
• 🚪 Easy exit - Type exit, quit, or q to leave

Pro Tips:

• All regular commands work: optimize, analyze, config, theme
• All flags work: --provider, --style, --iterations, --show-cost, etc.
• Guided wizards only show providers with configured API tokens
• Compare mode in wizards only shows providers you've set up
• Press Ctrl+C twice to exit immediately
• Perfect for batch operations and experimentation

1️⃣ Inline (Fastest!)

megabuff optimize "Write a function that validates emails"

2️⃣ From a File

megabuff optimize --file my-prompt.txt

3️⃣ Pipe It In

echo "Explain quantum computing" | megabuff optimize
cat prompt.txt | megabuff optimize

4️⃣ Interactive Prompt Input

megabuff optimize
# Type/paste your prompt, then Ctrl+D ✨

📤 Output Options

By default, MegaBuff does BOTH:

• ✅ Prints to stdout (for piping)
• ✅ Copies to clipboard (instant paste!)

Customize the output:

# Don't copy to clipboard
megabuff optimize "prompt" --no-copy

# Save to file (still copies by default)
megabuff optimize "prompt" --output result.txt

# Interactive before/after view
megabuff optimize "prompt" --interactive

# Save without clipboard
megabuff optimize "prompt" -o result.txt --no-copy

🎨 Optimization Styles

Tailor the optimization to your needs! Choose from 7 different styles:

# Balanced (default) - Well-rounded optimization
megabuff optimize "your prompt"

# Concise - Minimize tokens, maximize clarity
megabuff optimize "your prompt" --style concise

# Detailed - Add comprehensive context and examples
megabuff optimize "your prompt" --style detailed

# Technical - Perfect for code generation tasks
megabuff optimize "your prompt" --style technical

# Creative - Encourage imaginative outputs
megabuff optimize "your prompt" --style creative

# Formal - Professional, business-appropriate
megabuff optimize "your prompt" --style formal

# Casual - Friendly, conversational tone
megabuff optimize "your prompt" --style casual

Available Styles:

• balanced - Default, well-rounded optimization
• concise - Brief and to-the-point
• detailed - Comprehensive with examples
• technical - Precise technical terminology
• creative - Imaginative and flexible
• formal - Professional and structured
• casual - Conversational and approachable

🔧 Custom System Prompts

Ultimate control! Provide your own system prompt:

# Use a custom optimization strategy
megabuff optimize "your prompt" --system-prompt "Focus on making this prompt more concise while maintaining all technical details"

# Combine with any provider
megabuff optimize --provider anthropic "complex task" --system-prompt "Optimize for Claude's chain-of-thought reasoning"

Pro Tips:

• Custom prompts override all style and provider-specific optimizations
• Great for domain-specific optimization (legal, medical, scientific)
• Use to enforce company-specific prompt standards

🔄 Iterative Refinement

Progressive improvement! Run multiple optimization passes:

# Single pass (default)
megabuff optimize "your prompt"

# 3 iterations for progressive refinement
megabuff optimize "your prompt" --iterations 3

# Maximum refinement (5 passes)
megabuff optimize "complex prompt" --iterations 5

# Show output from each iteration (verbose mode)
megabuff optimize "your prompt" --iterations 3 --verbose

# Combine with styles and providers
megabuff optimize --iterations 3 --style detailed --provider anthropic "technical documentation"

How it works:

• Each iteration takes the previous output and optimizes it further
• Great for complex prompts that need multiple rounds of refinement
• Iterations are limited to 1-5 to balance quality and cost
• Progress is shown for each iteration with individual timing

Verbose mode (--verbose or -v):

• Shows the output from each iteration in real-time
• Helps you see how the prompt evolves with each pass
• Useful for understanding the optimization process
• Only displays iteration outputs when using multiple iterations

When to use multiple iterations:

• Complex, multi-part prompts
• Technical documentation that needs precision
• Creative writing prompts that benefit from layered refinement
• When you want the absolute best optimization possible

Analyze before optimizing (--analyze-first or -a):

• Get detailed analysis BEFORE optimization runs
• See exactly what will be improved
• Understand the prompt's weaknesses upfront
• Then automatically proceed with optimization

# Analyze first, then optimize
megabuff optimize "your prompt" --analyze-first
megabuff optimize "your prompt" -a --style technical

# Combine with iterations
megabuff optimize "your prompt" --analyze-first --iterations 3

🔍 Comparison Mode

Can't decide which provider to use? Test them all at once!

# Compare optimizations from all configured providers
megabuff optimize "your prompt" --compare

# Compare specific providers only
megabuff optimize "your prompt" --compare --providers openai,anthropic
megabuff optimize "your prompt" --compare --providers openai,anthropic,google

# Combine with styles and iterations
megabuff optimize "complex prompt" --compare --style technical
megabuff optimize "your prompt" --compare --iterations 3
megabuff optimize "your prompt" --compare --providers openai,deepseek --style concise

How it works:

• Runs optimization across all providers with configured API keys (or specified providers)
• Executes all providers in parallel for speed
• Shows side-by-side results with timing and statistics
• Requires at least 2 providers with configured API keys

Select specific providers:

• Use --providers flag with comma-separated provider names
• Example: --providers openai,anthropic,google
• Only tests the providers you specify (if they have API keys configured)
• Great for targeted comparisons

What you'll see:

• Individual results from each provider
• Duration and length statistics for each
• Average metrics across all successful providers
• Any errors or failures clearly marked

When to use comparison mode:

• Evaluating which provider works best for your use case
• Testing different AI approaches to the same problem
• Quality assurance - seeing multiple perspectives
• Finding the optimal provider for specific prompt types
• A/B testing between specific providers

Pro Tips:

• Configure multiple providers first using megabuff config set
• Results are NOT copied to clipboard (choose your favorite manually)
• Combine with --iterations for comprehensive testing
• Use --style to ensure consistent optimization approach across providers
• Use --providers to focus on specific providers you want to compare

💰 Cost Tracking & Estimation

Know before you spend! Estimate and track API costs for your operations with detailed per-token and per-model pricing information.

# Show cost estimate before running optimization
megabuff optimize "your prompt" --show-cost

# Only estimate cost without running (perfect for budgeting)
megabuff optimize "your prompt" --estimate-only

# Estimate with multiple iterations
megabuff optimize "your prompt" --iterations 3 --estimate-only

# Show cost for analysis
megabuff analyze "your prompt" --show-cost

# Estimate only for analysis
megabuff analyze "your prompt" --estimate-only

# Compare costs across providers (combine with --compare)
megabuff optimize "your prompt" --compare --show-cost

What you see with --show-cost:

1. Initial Cost Estimate (before operation runs):

• Model being used for the operation
• Detailed pricing breakdown:
  • Input price per 1M tokens and per individual token
  • Output price per 1M tokens and per individual token
• Estimated input tokens
• Estimated output tokens
• Estimated total cost in USD

2. Actual Cost Summary (after operation completes):

• Model used
• Detailed pricing breakdown (same as estimate)
• Actual input tokens used
• Actual output tokens used
• Actual total cost in USD
• Estimate accuracy percentage

3. Comparison Mode Pricing:

• Individual pricing details for each provider's model
• Total tokens used across all providers
• Total cost across all providers
• Average cost per provider
• Per-iteration cost breakdown (when using --iterations)

Example output:

💰 Cost Estimate
────────────────────────────────────────────────────────────────────────────────
   Model: claude-sonnet-4-5
   Pricing:
      Input:  $3.00/1M tokens ($0.000003000/token)
      Output: $15.00/1M tokens ($0.000015000/token)

   Input tokens: 512
   Output tokens (est): 614
   Estimated cost: $0.0025
────────────────────────────────────────────────────────────────────────────────

... [operation runs] ...

💰 Actual Cost
────────────────────────────────────────────────────────────────────────────────
   Model: claude-sonnet-4-5
   Pricing:
      Input:  $3.00/1M tokens ($0.000003000/token)
      Output: $15.00/1M tokens ($0.000015000/token)

   Optimization tokens: 512 in + 621 out
   Optimization cost: $0.0025
   Total cost: $0.0025
   Estimate accuracy: 98.9%
────────────────────────────────────────────────────────────────────────────────

How it works:

• Estimates token count based on your prompt text (~4 chars per token)
• Uses current pricing for each model/provider
• Accounts for system prompts and iterations
• Shows detailed cost breakdown BEFORE making any API calls
• Displays actual costs with pricing details AFTER operations complete
• Compares estimate vs actual for accuracy tracking

Pricing information (updated January 2025):

| Provider | Model | Input (per 1M) | Output (per 1M) | |----------|-------|----------------|-----------------| | OpenAI | gpt-5.2 | $2.50 | $10.00 | | | gpt-4o | $2.50 | $10.00 | | | gpt-4o-mini | $0.15 | $0.60 | | | gpt-4-turbo | $10.00 | $30.00 | | Anthropic | claude-opus-4-5 | $15.00 | $75.00 | | | claude-sonnet-4-5 | $3.00 | $15.00 | | | claude-sonnet-4-0 | $3.00 | $15.00 | | | claude-3-haiku | $0.25 | $1.25 | | Google | gemini-2.5-pro | $1.25 | $5.00 | | | gemini-2.5-flash | $0.075 | $0.30 | | | gemini-2.0-flash | $0.075 | $0.30 | | xAI | grok-beta | $5.00 | $15.00 | | | grok-vision-beta | $5.00 | $15.00 | | DeepSeek | deepseek-chat | $0.14 | $0.28 | | | deepseek-reasoner | $0.55 | $2.19 |

Prices are in USD per 1 million tokens

Cost calculation formula:

Cost = (input_tokens / 1,000,000 × input_price) + (output_tokens / 1,000,000 × output_price)

Pro Tips:

• Use --estimate-only to preview costs before committing
• Compare costs between providers to find the most economical option
• Most operations cost less than $0.01 with efficient models (gemini-flash, gpt-4o-mini, deepseek-chat)
• Use --show-cost to track cumulative spending in sessions
• Detailed pricing information helps you understand exactly where costs come from
• Per-token prices are shown with 9 decimal precision for accuracy
• In comparison mode, you'll see pricing details for each provider side-by-side

📊 Prompt Analysis

Not sure what's wrong with your prompt? Get detailed AI-powered feedback!

# Analyze a prompt and get comprehensive feedback
megabuff analyze "Write a function that does stuff"

# Analyze from a file
megabuff analyze --file my-prompt.txt

# Use a specific provider for analysis
megabuff analyze --provider anthropic "Explain quantum computing"

# Save analysis to a file
megabuff analyze "your prompt" --output analysis.txt

# Analyze without copying to clipboard
megabuff analyze "your prompt" --no-copy

What you get:

• Overall Assessment - Summary of the prompt's quality and purpose
• Strengths - 3-5 specific things the prompt does well
• Weaknesses & Issues - Problems, ambiguities, or areas for improvement
• Specific Suggestions - Actionable recommendations for enhancement
• Key Improvements - The most impactful changes to prioritize
• Clarity Score - Numeric rating (1-10) with justification

How it works:

• AI analyzes your prompt structure, clarity, and effectiveness
• Provides constructive, actionable feedback
• Identifies ambiguities and missing context
• Suggests concrete improvements
• Results are automatically copied to clipboard (unless --no-copy is used)

When to use prompt analysis:

• Before optimizing - understand what needs fixing first
• Learning prompt engineering - see what makes prompts effective
• Debugging poor AI responses - identify prompt issues
• Quality assurance - validate prompts meet standards
• Training others - explain good vs. bad prompt patterns

Pro Tips:

• Use analysis before optimization to understand what to focus on
• Different providers may highlight different aspects - try multiple!
• Save analyses to build a library of prompt engineering patterns
• Combine with optimize - analyze first, then optimize based on feedback

🎯 Examples

Real-world use cases to get you started:

# 🐚 Interactive shell mode (NEW!)
megabuff shell
# or: megabuff i
megabuff> optimize "Create a REST API" --style technical
megabuff> analyze "Write a chatbot"
megabuff> optimize --compare --providers openai,anthropic "Explain AI"
megabuff> theme set cyberpunk
megabuff> help
megabuff> exit

# 🚀 Quick optimization
megabuff optimize "Write code for user auth"

# 🧠 Use Claude for better reasoning
megabuff optimize --provider anthropic "Explain recursion"

# ✨ Try Gemini
megabuff optimize --provider google "Design a database schema"

# 📁 From a file with before/after comparison
megabuff optimize --file prompt.txt --interactive

# 🔄 Pipe and save
cat input.txt | megabuff optimize --output optimized.txt

# 🎯 Use a specific model
megabuff config model claude-sonnet-4-5
megabuff optimize "Explain quantum computing"

# 🎨 Make it pretty!
megabuff theme set cyberpunk           # Set theme
megabuff theme preview dracula         # Preview first
megabuff theme list                    # See all options

# 🎭 Use optimization styles
megabuff optimize --style technical "Create a REST API"
megabuff optimize --style concise "long rambling prompt here"
megabuff optimize --style creative "Write a story about AI"

# 🔬 Custom system prompts for specialized use
megabuff optimize "medical diagnosis criteria" --system-prompt "Optimize for medical professionals. Use precise medical terminology."

# 🔄 Iterative refinement for best results
megabuff optimize --iterations 3 "draft blog post intro"
megabuff optimize --iterations 5 --style detailed "complex technical spec"
megabuff optimize --iterations 3 --verbose "See how it evolves"

# 🔍 Analyze first, then optimize
megabuff optimize --analyze-first "Write a function that does stuff"
megabuff optimize -a --style technical "Create a REST API endpoint"
megabuff optimize --analyze-first --iterations 3 "Complex prompt that needs work"

# 🔍 Compare multiple providers side-by-side
megabuff optimize --compare "Create a REST API for user management"
megabuff optimize --compare --style technical "Explain quantum entanglement"
megabuff optimize --compare --iterations 3 "Write a product description"
megabuff optimize --compare --providers openai,anthropic "Which AI writes better code?"
megabuff optimize --compare --providers openai,anthropic,google --style concise "Short & sweet"

# 📊 Analyze prompts for detailed feedback
megabuff analyze "Write a function that does stuff"
megabuff analyze --provider anthropic "Explain how neural networks work"
megabuff analyze --file draft-prompt.txt --output analysis-report.txt
megabuff analyze "Create a chatbot for customer support" --provider openai

# 💰 Cost tracking and estimation
megabuff optimize "your prompt" --show-cost
megabuff optimize "your prompt" --estimate-only
megabuff optimize "your prompt" --iterations 5 --estimate-only
megabuff analyze "your prompt" --show-cost
megabuff optimize --compare --providers openai,anthropic --show-cost "Cost comparison"

# 🔧 Power user combos
megabuff optimize --file long-prompt.txt --provider anthropic -o result.txt --interactive
megabuff optimize --provider deepseek --model deepseek-reasoner --style detailed "Complex reasoning task"
megabuff optimize --iterations 3 --style technical --provider anthropic "API documentation"
megabuff optimize --compare --providers openai,deepseek --style concise --iterations 2 "Head-to-head test"

# 💡 Analysis + Optimization workflow
megabuff analyze "my prompt" --output analysis.txt    # Understand issues first
megabuff optimize "my prompt" --style technical       # Then optimize based on feedback

🔧 How It Works

The Magic Behind MegaBuff ✨

MegaBuff uses state-of-the-art AI to transform your prompts:

🎯 What Gets Optimized?

1. Clarity - Removes ambiguity and vague instructions
2. Context - Adds missing details that improve results
3. Structure - Organizes information logically
4. Format - Specifies expected output clearly
5. Specificity - Makes requests actionable and precise

🤖 Choose Your AI

| Provider | Best For | Default Model | |----------|----------|---------------| | 🤖 OpenAI | Fast & economical | GPT-4o-mini | | 🧠 Anthropic | Deep reasoning | Claude Sonnet 4.5 | | ✨ Google | Efficient & quality | Gemini 1.5 Flash |

Switch providers anytime with --provider or set your favorite as default!

🆚 VS Code Integration

Use MegaBuff right inside VS Code! 🎉

⭐ Option 1: VS Code Extension (Full Experience)

Install and develop the MegaBuff VS Code extension:

cd ../megabuff-vscode
npm install
npm run compile

Then press F5 to launch the extension in debug mode.

Features:

• Right-click context menu for selected text
• Command palette integration
• Keyboard shortcuts (Ctrl+Shift+Alt+O)
• Diff view for before/after comparison
• API key management UI
• Status bar integration

See megabuff-vscode/README.md for more details.

🚀 Option 2: VS Code Tasks (Quick Setup)

Pre-configured tasks for instant productivity:

1. Select text
2. Ctrl+Shift+P → "Tasks: Run Task"
3. Pick a MegaBuff task

Available Tasks:

• ✨ Optimize Selected Text
• 📄 Optimize Current File
• 🔍 Optimize (Interactive View)
• ⚙️ Configure API Key

💻 Option 3: Terminal Integration

Just use the integrated terminal:

pbpaste | megabuff optimize  # macOS
xclip -o | megabuff optimize  # Linux

🤝 Contributing

We'd love your help making MegaBuff better!

Found a bug? Have an idea? Open an issue:

👉 GitHub Issues

📄 License

MegaBuff is open source! Licensed under AGPL-3.0

What this means:

• ✅ Free to use - For any purpose
• ✅ Free to modify - Make it your own
• ✅ Free to share - Spread the love
• 🔓 Open source required - Derivative works must stay open
• 🌐 Service transparency - If you run it as a service, share the code

Read the full license

👨‍💻 Development

Want to contribute or build from source?

🔧 Local Development

# Clone the repo
git clone https://github.com/thesupermegabuff/megabuff-cli.git
cd megabuff-cli

# Install dependencies
npm install

# Run in dev mode (note the -- separator before arguments)
npm run dev -- optimize "Your prompt"

# With flags (always use -- to pass arguments to the script)
npm run dev -- optimize "Your prompt" --compare --providers openai,anthropic
npm run dev -- optimize "Your prompt" --style technical --iterations 3
npm run dev -- optimize "Your prompt" --compare --providers google,openai --style concise --iterations 3
npm run dev -- optimize "Your prompt" --compare --providers google,openai --style concise --iterations 2 --verbose

# Or use tsx directly (no -- needed)
npx tsx src/index.ts optimize "Your prompt"
npx tsx src/index.ts optimize "Your prompt" --compare --providers openai,anthropic

📦 Build & Publish

Building the project:

npm run build  # Compiles TypeScript to dist/

Version and build:

npm version minor  # Bump version (patch/minor/major)
npm run build      # Compile TypeScript

This compiles TypeScript to JavaScript in the dist/ folder.

Test the package locally (optional but recommended):

npm pack
# This creates a .tgz file you can inspect

Test locally with npm link:

npm link  # Install as global command
megabuff optimize "Test it out!"

Publish to npm:

npm publish

The prepublishOnly script will automatically run npm run build before publishing.

📚 Publishing to npm

🎯 First-time Setup

1. Create an npm account
2. Login: npm login
3. Update package.json with your details
4. Check name availability: npm search megabuff

🚀 Publishing Process

# 1. Commit changes
git add .
git commit -m "Release preparation"

# 2. Bump version (patch/minor/major)
npm version minor

# 3. Publish!
npm login
npm publish

# 4. Push tags
git push && git push --tags

📦 What Gets Published

• ✅ Compiled JavaScript (dist/)
• ✅ Documentation
• ✅ Package metadata
• ❌ TypeScript source
• ❌ Dev dependencies

🎉 After Publishing

Users install with:

npm install -g megabuff

Available at: npmjs.com/package/megabuff

Made with ❤️ by the MegaBuff Team

⭐ Star us on GitHub!

🐛 Report Issues • 💡 Request Features