Reactive Agents

Automatically optimize your AI agents based on performance with OpenAI API compatibility.

Reactive Agents is a self-optimizing AI agent platform that automatically improves your agents by adjusting hyperparameters and system prompts based on their performance. Simply point your existing application to Reactive Agents instead of your AI provider, and it handles the optimization for you.

Key Features:

• 🔄 Automatic optimization - Continuously improves agents based on real-world performance
• 🎯 Smart tuning - Adjusts LLM hyperparameters and system prompts automatically
• 🔌 OpenAI API compatible - Works as a drop-in replacement for applications using the OpenAI API
• 🌐 Multi-provider support - Route to OpenAI, Anthropic, Google, and other providers
• 📊 Performance tracking - Monitor agent performance and improvements over time

How it works:

1. Change your API URL from api.openai.com to your Reactive Agents instance
2. Reactive Agents proxies requests to your chosen AI provider
3. Performance is tracked and agents are automatically optimized
4. Your application benefits from continuously improving AI agents

Quick Start:

1. Create a .env file:

   # For now, we use GPT-5 for the system prompt reflection and GPT-5-mini for evaluations.
   OPENAI_API_KEY=your-openai-key

2. Start the application:

   docker-compose up

   The application will be available at http://localhost:3000.

3. See examples/openai-lib/chat-complete-api-structured-output.ts for a detailed example of how to use Reactive Agents with OpenAI's Chat Completion API. You will need to set up a calendar_event_planner agent with a generate skill in the UI.

What's included:

• Complete backend infrastructure
• Reactive Agents application
• All necessary setup handled automatically

For future production deployment guides and best practices, visit our documentation.

Check out the examples/ directory for sample implementations showing how to integrate Reactive Agents with your applications.

• Documentation - Complete guides and API reference
• Contributing Guide
• Notice

Made with ❤️ by the Reactive Agents team