Building crypto portfolios with agentic AI

The rapid growth of crypto markets is increasing investment opportunities, but also volatility risks. There is a need to automate portfolio construction to rapidly adapt investment decisions to rapidly changing markets. To fill the gap, in this paper we propose a multi-agent system designed to autonomously construct and evaluate crypto-asset allocations, in terms of returns and risks. We exemplify the methodology by its application to daily frequency return data of the ten most capitalized cryptocurrencies, from 2020 to 2025. Specifically, we compare two automated investment strategies: a static strategy and a rolling-window optimization strategy, both based on the optimisation of evaluation metrics such as the Expected Return and Sharpe ratio. The results show that the dynamic optimization strategy achieves better performance in terms of risk-adjusted returns, both in-sample and out-of-sample. This highlights the benefits of multi-agent systems in portfolio management, particularly in volatile markets such as cryptocurrency markets.