Multi-LLM trading harness with live leaderboard on Alpaca paper trades (github.com)

A new standalone harness called **1rok** has been announced, enabling the execution of portfolio-construction agents from various large language model (LLM) providers, including OpenAI, Anthropic, and others, against a common financial toolset. This system utilizes live data from sources like Alpaca and Yahoo Finance to facilitate paper trading and track performance through a live leaderboard. By supporting multiple LLMs within a unified environment, 1rok enhances accessibility and competition among different AI models in financial decision-making. Significantly, the 1rok framework features a two-stage pipeline that separates portfolio construction from order execution, ensuring that the “run” command creates a JSON artifact for portfolio proposals, while the “execute” command handles the placement of orders, typically in a simulated environment. This architecture incorporates various specialist agents—including orchestrators and analysts—each designed to evaluate different market factors, from sentiment and risk to macroeconomic indicators. By allowing flexible integrations and provider-agnostic schemas for tool-calling, this initiative not only promotes advancements in AI-driven trading strategies but also offers a platform for comparative analysis of AI model performance in real-time financial scenarios.