This paper proposes an Agent-based Simulated Financial Market (ASFM) framework that uses large language models (LLMs) to simulate financial market behavior. ASFM constructs a simulated stock market with a real order-matching system and employs LLM-based agents as stock traders. These agents have profiles, observation modules, and tool-learning based action modules to make decisions aligned with their trading strategies. The framework simulates various financial market scenarios, including interest rate changes and inflation shocks, and demonstrates that the simulated market aligns with real-world economic research findings. The ASFM framework includes a simulated stock market with multiple listed companies and an order-matching system that mimics real-world trading mechanisms. It also includes LLM-based trading agents with diverse investment strategies, such as value investing, institutional investing, contrarian investing, and aggressive investing. These agents observe market trends, receive economic news, and use tool learning to perform stock trading in the simulated market. The paper evaluates the effectiveness of ASFM through various experiments, including the impact of interest rate cuts and inflation shocks on the stock market. The results show that ASFM accurately simulates real-world market responses and aligns with economic research findings. Additionally, the framework explores the impact of trader behavior bias and large trader influence on market dynamics, demonstrating consistency with preliminary economic research. ASFM provides a new paradigm for economic research by integrating LLMs with economic and financial operations. It offers a scalable and cost-effective alternative to traditional experimental economics methods, enabling the simulation of complex financial market behaviors. The framework has the potential to revolutionize economic research by providing a more accurate and dynamic simulation of financial markets.This paper proposes an Agent-based Simulated Financial Market (ASFM) framework that uses large language models (LLMs) to simulate financial market behavior. ASFM constructs a simulated stock market with a real order-matching system and employs LLM-based agents as stock traders. These agents have profiles, observation modules, and tool-learning based action modules to make decisions aligned with their trading strategies. The framework simulates various financial market scenarios, including interest rate changes and inflation shocks, and demonstrates that the simulated market aligns with real-world economic research findings. The ASFM framework includes a simulated stock market with multiple listed companies and an order-matching system that mimics real-world trading mechanisms. It also includes LLM-based trading agents with diverse investment strategies, such as value investing, institutional investing, contrarian investing, and aggressive investing. These agents observe market trends, receive economic news, and use tool learning to perform stock trading in the simulated market. The paper evaluates the effectiveness of ASFM through various experiments, including the impact of interest rate cuts and inflation shocks on the stock market. The results show that ASFM accurately simulates real-world market responses and aligns with economic research findings. Additionally, the framework explores the impact of trader behavior bias and large trader influence on market dynamics, demonstrating consistency with preliminary economic research. ASFM provides a new paradigm for economic research by integrating LLMs with economic and financial operations. It offers a scalable and cost-effective alternative to traditional experimental economics methods, enabling the simulation of complex financial market behaviors. The framework has the potential to revolutionize economic research by providing a more accurate and dynamic simulation of financial markets.