Use LLM agents to debate a trade idea

They built a multi-agent setup that weighs a stock from several angles, then routes a recommendation through review stages before any simulated order is sent.

Done with Custom agent· Advanced

What they did
They installed the project, added an API key for an LLM provider, and chose a ticker and date either in the command line or through a Python graph call. Different agents handled company basics, market mood, recent headlines, and chart signals, then bullish and bearish roles argued over the idea. After that, a trader agent made a call, risk reviewers checked it, and a portfolio manager either allowed it or stopped it before anything reached a mock exchange.
How it went
The project did not publish benchmark scores. It also warned that results can change from run to run, and backtest performance may differ from any reported figures, so it is better viewed as a research framework than a fixed-return system.
Tips
Use a non-reasoning model if you want more stable behavior, since reasoning models ignore the temperature setting. Keep it in research mode and do not link it to a live brokerage.

Try it yourself with Claude Code

Build a small multi-agent research workflow in this folder for [ticker or tickers]. Use separate agents for fundamentals, news, technical signals, sentiment, bullish and bearish review, and risk checks, then have a final reviewer either approve or reject one written trade idea per ticker. Base the work only on [data source], explain the reasoning in plain language, and stop after saving the report for [ticker]. Before any irreversible action such as sending, submitting, deleting, or placing an order, pause and ask me to confirm. Do not connect to a real broker or make live trades.

Fill in the [brackets]. Agents can make mistakes, so review anything before it is sent, paid or deleted.

Read the original on github.com ↗

Get 3 agent uses like this every morning

With the day's top AI news. Free, and one click to unsubscribe.