AI Digital Twins trained on real user behavior, not LLM persona prompts.
Agentic World captures real user behavior from web sessions, clusters users into behavioral demographics, and deploys AI agents that autonomously QA-test new features before real users see them.
Chrome Extension → FastAPI Backend → SQLite Storage
→ NeMo Data Designer (synthetic augmentation)
→ Mistral/NVIDIA AI Pipeline
→ Playwright Agent Runner
→ Next.js Dashboard
- Mistral Small 3.1 — Real-time agent decision loops (low latency click sequence replay)
- Codestral — Auto-generates DOM interaction scripts from behavioral recordings
- Mistral Medium 3 — Behavioral reasoning layer (understands why users hesitate)
- Llama-3.1-Nemotron-49B — Demographic inference from behavioral signals
- NeMo Guardrails — Prevents destructive actions during QA testing
- NeMo Data Designer — Generates synthetic behavioral session data at scale
agentic-world/
├── chrome-extension/ # Manifest V3 Chrome extension for behavior capture
├── backend/ # FastAPI server with AI pipeline
└── dashboard/ # Next.js business analytics dashboard
- Python 3.11+
- Node.js 18+
- API keys for Mistral and NVIDIA
cd backend
python -m venv venv
source venv/bin/activate
pip install -r requirements.txt
playwright install chromium
cp .env.example .env
# Edit .env with your API keys
uvicorn main:app --reload --port 8000cd dashboard
npm install
npm run dev
# Opens at http://localhost:3000- Open
chrome://extensions/in Chrome - Enable "Developer mode"
- Click "Load unpacked" → select the
chrome-extension/folder - Browse any website — the extension captures your behavior
MISTRAL_API_KEY=your_mistral_api_key
NVIDIA_API_KEY=your_nvidia_api_key- Real-time behavior capture — mouse movements, clicks, scroll velocity, hover duration, hesitation patterns, rage clicks
- Behavioral clustering — automatically discovers user archetypes (Cautious Explorer, Power User, Scanner, etc.)
- Synthetic data augmentation — NeMo Data Designer generates training data at scale
- Autonomous QA agents — Playwright-powered digital twins that test your product
- A/B test simulation — synthetic A/B results before real users touch your product
- Insights dashboard — heatmaps, funnels, hesitation hotspots, AI-generated recommendations