← Back
Evidence source 5874Spot Checked

Orchestration Framework for Financial Agents: From Algorithmic Trading to Agentic Trading

arXiv2025-12-01Paper
Executive summary

A new framework for financial trading replaces traditional algorithms with a network of specialized AI agents, each handling tasks like planning, risk management, and execution. These agents use structured communication and shared memory for transparency and reproducibility. The system outperformed benchmarks, achieving a 20.42% return and a Sharpe ratio of 2.63 in stocks, and 8.39% in Bitcoin trading. The approach aims to make advanced trading tools public, though limited backtesting and market coverage remain concerns.

What it examines

This paper introduces an orchestration framework for financial agents, mapping each part of traditional algorithmic trading to specialized AI agents. The goal is to make advanced trading tools accessible to everyone, using multi-agent systems and large language models to handle data, signals, risk, portfolios, and execution.

What it concludes

The framework outperformed standard benchmarks in both stock and crypto trading, showing better risk control and returns. Potential applications include automated trading, portfolio management, and risk analysis. Future work will expand to more markets, longer timeframes, and improved agent communication and memory for safer, more adaptive trading systems.

Extracted from this source

Evidence objects

Evidence 634382% extraction confidence
A new AI-powered framework, 'agentic trading,' replaces traditional algorithmic trading by mapping key financial tasksplanning, risk, execution, auditto specialized agents, ensuring transparency, reproducibility, and modularity through structured protocols and shared memory.

key_findings bullet 1 · key_findings · validation V0

Evidence 634482% extraction confidence
The system outperformed benchmarks: in stock trading, it achieved a 20.42% return and Sharpe ratio of 2.63, beating the S&P 500; in high-frequency Bitcoin trading, it earned 8.39% versus 3.8% for Buy-and-Hold, with lower risk.

key_findings bullet 2 · key_findings · validation V0

Evidence 634582% extraction confidence
Leveraging large language models for reasoning and strict safeguards against data leakage, the framework democratizes advanced trading, but short backtest periods and limited market coverage highlight the need for broader, long-term validation.

key_findings bullet 3 · key_findings · validation V0

Evidence 634682% extraction confidence
This paper introduces FinAgent, a novel framework mapping traditional trading components to agentic architectures using LLMs and multi-agent orchestration. Its originality lies in explicit communication protocols, modular agent pools, and practical implementation, including open-source code. Demonstrated on equity and crypto tasks, it offers compelling, timely advances as the field shifts paradigms.

key_findings bullet 4 · key_findings · validation V0

Raw abstract and provenance

Abstract: The financial market is a mission-critical playground for AI agents due to its temporal dynamics and low signal-to-noise ratio. Building an effective algorithmic trading system may require a professional team to develop and test over the years. In this paper, we propose an orchestration framework for financial agents, which aims to democratize financial intelligence to the general public. We map e… ▽ More The financial market is a mission-critical playground for AI agents due to its temporal dynamics and low signal-to-noise ratio. Building an effective algorithmic trading system may require a professional team to develop and test over the years. In this paper, we propose an orchestration framework for financial agents, which aims to democratize financial intelligence to the general public. We map each component of the traditional algorithmic trading system to agents, including planner, orchestrator, alpha agents, risk agents, portfolio agents, backtest agents, execution agents, audit agents, and memory agent. We present two in-house trading examples. For the stock trading task (hourly data from 04/2024 to 12/2024), our approach achieved a return of $20.42\%$, a Sharpe ratio of 2.63, and a maximum drawdown of $-3.59\%$, while the S&P 500 index yielded a return of $15.97\%$. For the BTC trading task (minute data from 27/07/2025 to 13/08/2025), our approach achieved a return of $8.39\%$, a Sharpe ratio of $0.38$, and a maximum drawdown of $-2.80\%$, whereas the BTC price increased by $3.80\%$. Our code is available on \href{https://github.com/Open-Finance-Lab/AgenticTrading}{GitHub}. △ Less

Source row: 1523 · abstract type: unknown