Multi-Agent Paper Trading Bot
A multi-agent trading system using Claude and a retrieval layer to reason about paper trades on live market data.
Trading Application · Planned
Problem
Most trading bots follow rigid, hand-coded rules and can't explain their reasoning. Testing a more adaptive, LLM-driven strategy safely requires a paper-trading environment where agents can reason, retrieve context, and act without risking real capital.
Approach
A NestJS backend orchestrates specialized agents (market analysis, risk, execution) built on the Claude API, backed by a Next.js dashboard. Trades run against Alpaca's paper trading API, and a RAG layer (PostgreSQL with pgvector) gives the agents relevant historical context before each decision.
Status
Stack decided, no code written yet. Development has not started.
Stack
- NestJS
- Next.js
- Alpaca API
- Claude API
- PostgreSQL
- pgvector