An interactive web demo showcasing SCX.ai's Agentic Context Engineering (ACE) technology for self-improving AI systems.
🌐 Visit SCX.ai | 📄 Research Paper
SCX.ai brings you Agentic Context Engineering — a breakthrough framework where AI models evolve their own playbooks through a modular process of:
- Generator - Produces reasoning trajectories using the playbook
- Reflector - Analyzes outputs and tags strategies as helpful/harmful
- Curator - Updates the playbook with delta context items
- 86.9% Faster Adaptation - Lower latency through incremental delta updates
- +10.6% Accuracy Gain - Preserves domain-specific knowledge
- Zero Retraining - Self-improves without labeled data or fine-tuning
- Node.js 18+
- SCX.ai API key
# Clone the repo
git clone https://github.com/Monty1122/ace-demo.git
cd ace-demo
# Install dependencies
npm install
# Set up environment variables
cp .env.example .env.local
# Edit .env.local with your SCX.ai API key
# Run development server
npm run devCreate a .env.local file:
LLM_API_URL=https://api.scx.ai/v1
LLM_API_KEY=your-scx-ai-api-key
LLM_MODEL=DeepSeek-V3.1ace-demo/
├── app/
│ ├── page.tsx # Landing page with SCX.ai branding
│ ├── demo/
│ │ └── page.tsx # Interactive ACE demo
│ ├── api/
│ │ └── ace/
│ │ └── route.ts # ACE pipeline API
│ ├── layout.tsx
│ └── globals.css
├── public/
│ ├── scx-logo-long.png
│ └── scx-logo-square.png
├── package.json
└── README.md
- Interactive Demo - Run the three-agent pipeline on sample problems
- Live Playbook View - Watch the playbook evolve with helpful/harmful counters
- Agent Visualization - Animated pipeline showing Generator → Reflector → Curator
- Sample Problems - Pre-loaded financial NER and calculation examples
- Custom Questions - Try your own questions with SCX.ai's AI
- SCX.ai Website - Learn more about SCX.ai
- ACE Research Paper - Read the academic paper
- Original ACE Repository - Open source implementation
SCX.ai is revolutionizing how AI systems learn and adapt. Our platform enables enterprises to deploy AI that continuously improves through real-world usage — no retraining needed.
The Future of Self-Improving AI
Powered by SCX.ai • Based on research from Stanford & SambaNova