Prompt Optimizer is a powerful AI prompt optimization tool that helps you write better AI prompts and improve the quality of AI outputs. It supports four usage methods: web application, desktop application, Chrome extension, and Docker deployment.

Prompts can start from manual writing, templates, local imports, or sources such as Prompt Garden. Prompt Optimizer is where those prompts are optimized, tested, evaluated, and saved as reusable prompt assets.

1. Hard-Nosed Reviewer: Turn Agreement into Useful Critique

Starting from a minimal English role prompt, optimization pushes a small model away from generic pushback and toward a clearer, more structured review that surfaces weak assumptions, evidence gaps, and concrete revision advice.

2. Marketplace Bargaining Reply: Let Variables Change the Strategy

With a single reusable prompt template, you can swap in item details, price anchors, buyer offers, tone, and negotiation goals for different marketplace conversations. After optimization, the same small model does a better job turning those variables into a clearer, more transaction-ready reply instead of a generic helper-style response.

3. Text-to-Image: Optimize a One-Line Idea into a More Directable Key Visual Prompt

This is not just prompt expansion. Starting from a vague one-line idea, Prompt Optimizer adds clearer subject cues, spatial relationships, and mood anchors. The left side is simply “a floating library in the night sky,” while the optimized version gives the model a more directed fantasy composition that feels closer to a reusable key visual than a lucky generic image.

• 🎯 Intelligent Optimization: One-click prompt optimization with multi-round iterative improvements to enhance AI response accuracy
• 📝 Dual Mode Optimization: Support for both system prompt optimization and user prompt optimization to meet different usage scenarios
• 🔄 Analysis and Compare Evaluation: Supports analysis, single-result evaluation, and multi-result compare evaluation to help determine whether a prompt has truly improved
• 🤖 Multi-model Integration: Support for mainstream AI models including OpenAI, Gemini, DeepSeek, Grok, Zhipu AI, SiliconFlow, MiniMax, etc.
• 🖼️ Image Generation: Support for Text-to-Image (T2I), Image-to-Image (I2I), and Multi-Image generation with models like Gemini, Seedream, Grok
• 🌱 Prompt Sources: Start from manual writing, templates, local imports, or Prompt Garden import codes
• ⭐ Smart Favorites: Resource-aware prompt assets with version history, reproducible examples, media support, source binding, and workspace application
• 📊 Advanced Testing Mode: Context variable management, multi-turn conversation testing, Function Calling support
• 🔒 Secure Architecture: Pure client-side processing with direct data interaction with AI service providers, bypassing intermediate servers
• 📱 Multi-platform Support: Available as web application, desktop application, Chrome extension, and Docker deployment
• 🔐 Access Control: Password protection feature for secure deployment
• 🧩 MCP Protocol Support: Supports Model Context Protocol (MCP), enabling integration with MCP-compatible AI applications like Claude Desktop

• 🖼️ Text-to-Image (T2I): Generate images from text prompts
• 🎨 Image-to-Image (I2I): Transform and optimize images based on local files
• 🖼️ Multi-Image Generation: Use multiple input images to constrain subject relationships, sequential semantics, and final generation goals
• 🔌 Multi-model Support: Integrated with mainstream image generation models like Gemini, Seedream, Grok
• ⚙️ Model Parameters: Support model-specific parameter configuration (size, style, etc.)
• 📥 Preview & Download: Real-time preview of generated results with download support
• 🔄 Style Transfer: Learn style, composition, and color from reference images

• 🌱 Optional Prompt Sources: Bring prompts from manual writing, templates, local files, or Prompt Garden
• 📥 Import & Collect: Import prompts with metadata, media, examples, and source binding when available
• ⭐ Resource-aware Assets: Save stable prompts as reusable favorites with version history
• 🔗 Source Binding: Track prompt origins and maintain reproducible examples without requiring a specific source
• 📦 Complete Backup: Export and import favorites with all referenced resources

• 📊 Context Variable Management: Custom variables, batch replacement, variable preview
• 💬 Multi-turn Conversation Testing: Simulate real conversation scenarios to test prompt performance in multi-turn interactions
• 🛠️ Function Calling Support: Function Calling integration with support for OpenAI and Gemini tool calling
• 🔍 Analysis and Evaluation Pipeline: Supports analysis, evaluation, compare evaluation, and evaluation-driven smart rewrite in text modes

For detailed usage instructions, please refer to the Image Mode Documentation

Direct access: https://prompt.always200.com

The online app stores its app data in your browser. Model requests go directly to the provider you configure. For your first run, connect one working text model in Model Manager, then follow the first-time tutorial to improve and copy a prompt.

Method 1: One-click deployment to your own Vercel:

Method 2: Fork the project and import to Vercel (Recommended):

• First fork the project to your GitHub account
• Then import the project to Vercel
• This allows tracking of source project updates for easy syncing of new features and fixes
• Configure environment variables:
  • ACCESS_PASSWORD: Set access password to enable access restriction
  • VITE_OPENAI_API_KEY etc.: Optional private-deployment model settings. Do not preconfigure API keys on public frontend deployments because VITE_* values are exposed in browser assets.

For more detailed deployment steps and important notes, please check:

• Vercel Deployment Guide
• Cloudflare Deployment Guide

Use the Deploy to Cloudflare button for the quickest public-repository setup. It creates a repository under your GitHub/GitLab account and deploys with Workers Builds. For private repositories or stricter repository-access control, import your own repository manually; keep the default deploy commands and clear the build command if Cloudflare auto-fills pnpm run build, because wrangler.jsonc builds the Web frontend and publishes packages/web/dist as static assets.

For access control and analytics on Cloudflare, configure Cloudflare Access and Cloudflare Web Analytics in the Cloudflare dashboard. No frontend dependency or application-code change is required.

Download the latest version from GitHub Releases. We provide both installer and archive formats for each platform.

• Installer (Recommended): Such as *.exe, *.dmg, *.AppImage, etc. Strongly recommended as it supports automatic updates.
• Archive: Such as *.zip. Extract and use, but cannot auto-update.

Core Advantages of Desktop Application:

• ✅ No CORS Limitations: As a native desktop application, it completely eliminates browser Cross-Origin Resource Sharing (CORS) issues. This means you can directly connect to any AI service provider's API, including locally deployed Ollama or commercial APIs with strict security policies, for the most complete and stable functional experience.
• ✅ Automatic Updates: Versions installed through installers (like .exe, .dmg) can automatically check and update to the latest version.
• ✅ Independent Operation: No browser dependency, providing faster response and better performance.

1. Install from Chrome Web Store (may not be the latest version due to approval delays): Chrome Web Store
2. Click the icon to open the Prompt Optimizer

</details>

### 6. Docker Compose Deployment
<details>
<summary>Click to view Docker Compose deployment steps</summary>
```bash
# 1. Clone the repository
git clone https://github.com/linshenkx/prompt-optimizer.git
cd prompt-optimizer

# 2. Create .env file for API keys and authentication
cat > .env << EOF
# API Key Configuration
VITE_OPENAI_API_KEY=your_openai_api_key
VITE_GEMINI_API_KEY=your_gemini_api_key
VITE_DEEPSEEK_API_KEY=your_deepseek_api_key
VITE_GROK_API_KEY=your_xai_api_key
VITE_ZHIPU_API_KEY=your_zhipu_api_key
VITE_SILICONFLOW_API_KEY=your_siliconflow_api_key

# Basic Authentication (Password Protection)
ACCESS_USERNAME=your_username  # Optional, defaults to "admin"
ACCESS_PASSWORD=your_password  # Set access password
EOF

# Because the compose file is under docker/, pass the root .env explicitly.

# 3. Start the service
docker compose --env-file .env -f docker/docker-compose.yml up -d

# 4. View logs
docker compose --env-file .env -f docker/docker-compose.yml logs -f

# 5. Access the service
Web Interface: http://localhost:8081
MCP Server: http://localhost:8081/mcp

You can also directly edit the docker/docker-compose.yml file to customize your configuration:

services:
  prompt-optimizer:
    # Use Docker Hub image
    image: linshen/prompt-optimizer:latest
    container_name: prompt-optimizer
    restart: unless-stopped
    ports:
      - "8081:80"  # Web application port (MCP server accessible via /mcp path)
    environment:
      - VITE_OPENAI_API_KEY=your_openai_key
      - VITE_GEMINI_API_KEY=your_gemini_key
      - VITE_GROK_API_KEY=your_xai_key
      # Access Control (Optional)
      - ACCESS_USERNAME=admin
      - ACCESS_PASSWORD=your_password

# MCP Server Configuration
MCP_DEFAULT_MODEL_PROVIDER=openai  # Options: openai, gemini, anthropic, deepseek, grok, siliconflow, zhipu, dashscope, openrouter, modelscope, custom
MCP_LOG_LEVEL=info                 # Log level

docker run -d -p 8081:80 \
  -e VITE_OPENAI_API_KEY=your-openai-key \
  -e MCP_DEFAULT_MODEL_PROVIDER=openai \
  --name prompt-optimizer \
  linshen/prompt-optimizer

{
  "services": [
    {
      "name": "Prompt Optimizer",
      "url": "http://localhost:8081/mcp"
    }
  ]
}

-e VITE_OPENAI_API_KEY=your_key
-e VITE_GEMINI_API_KEY=your_key
-e VITE_DEEPSEEK_API_KEY=your_key
-e VITE_GROK_API_KEY=your_key
-e VITE_ZHIPU_API_KEY=your_key
-e VITE_SILICONFLOW_API_KEY=your_key

# Multiple Custom Models Configuration (Unlimited Quantity)
-e VITE_CUSTOM_API_KEY_ollama=dummy_key
-e VITE_CUSTOM_API_BASE_URL_ollama=http://localhost:11434/v1
-e VITE_CUSTOM_API_MODEL_ollama=qwen2.5:7b

For detailed documentation, see Development Documentation

# 1. Clone the project
git clone https://github.com/linshenkx/prompt-optimizer.git
cd prompt-optimizer

# 2. Install dependencies
pnpm install

# 3. Start development server
pnpm dev               # Main development command: build core/ui and run web app
pnpm dev:web          # Run web app only
pnpm dev:fresh        # Complete reset and restart development environment

• Basic feature development
• Web application release
• Chrome extension release
• Internationalization support
• Support for system prompt optimization and user prompt optimization
• Desktop application release
• MCP service release
• Advanced mode: Variable management, context testing, function calling
• Image generation: Text-to-Image (T2I) and Image-to-Image (I2I) support
• Prompt favorites and template management
• Support for workspace/project management

For detailed project status, see Project Status Document

• Documentation Index - Index of all documentation
• Technical Development Guide - Technology stack and development specifications
• LLM Parameters Configuration Guide - Detailed guide for advanced LLM parameter configuration
• Project Structure - Detailed project structure description
• Project Status - Current progress and plans
• Product Requirements - Product requirements document
• Vercel Deployment Guide - Detailed instructions for Vercel deployment
• Cloudflare Deployment Guide - Web frontend deployment on Cloudflare Workers / Pages

# For installed applications
xattr -rd com.apple.quarantine /Applications/PromptOptimizer.app

# For downloaded .dmg files (run before installation)
xattr -rd com.apple.quarantine ~/Downloads/PromptOptimizer-*.dmg

Thanks to all the developers who have contributed to this project!

This project was partly inspired by LangGPT in prompt engineering and structured prompt design. Thanks to the LangGPT project and community for their open-source sharing and continued exploration.

This project is licensed under AGPL-3.0.

In simple terms: You can freely use, modify, and commercialize this project, but if you turn it into a website or service for others, you must share your source code.

If this project is helpful to you, please consider giving it a Star ⭐️

• Submit an Issue
• Create a Pull Request
• Join the discussion group