Sentinel – Safe Agentic Paper-Trading Assistant
An auditable AI trading agent that paper-trades on its own within hard risk limits, with a kill switch and a plain-language journal.
Built at AgentX Hackathon 2026
Team Chaithu1705: Chaithanya Kumar, gandavaram naveena
Built with Python, FastAPI, pandas, microservices, HTML/JS dashboard, optional Grok LLM news scorer
About the project
Retail traders and trading bots break their own risk rules. Sentinel is an agentic assistant that runs the whole decision loop safely on paper. Market data comes in, a strategy is chosen by backtest score, and news can veto or shrink a trade. A risk guard then enforces a 5% max position, a 2% stop, a minimum reward-to-risk of 1.5 and a 3% daily loss cap before the paper broker fills anything. A kill switch halts trading instantly, and every decision is written to a plain-language journal. It's built as seven small services. It places no real orders and makes no profit claims; the goal is a trustworthy, auditable agent.