TauricResearch / TradingAgents Multi-agent LLM framework

Multi-agents LLM financial trading framework

One model agreeing with itself is not a desk.

TradingAgents is a multi-agent LLM financial trading framework. Separate agents hold separate roles, so a position has to survive disagreement before it becomes a decision.

Roles at the tableIllustrative
Research reads the filings and the news flow stancebullish
Risk asks what this costs when it is wrong stancebearish
Trader sizes it against what the book already holds stancehold
Decision taken after the disagreement, not before it outputlogged

Role names and stances are illustrative. The project describes itself as a multi-agent LLM financial trading framework. Nothing here is advice or a result.

01

Roles, not personas

A research seat and a risk seat want different things, and that conflict is the reason to run more than one agent.

02

Disagreement is the product

A decision that no agent argued against tells you less than one that survived an argument.

03

A framework, not a strategy

The repository gives the structure for running the desk. What it trades and how well remains yours to answer.

TauricResearch/TradingAgents Design study by Goose. Not affiliated with the project.