Are Autonomous Agents the Inevitable Future of Decentralized Finance? A Review and Forward Outlook

Decentralized finance (DeFi) is expanding rapidly with the exponential increase in several adoption metrics. However, most user activity and decisions in DeFi still rely on short-term information, which exposes participants to volatility, manipulation, and security risk. Autonomous agents offer a data-driven approach to mitigate those frictions. This study conducts a systematic literature review (SLR) to map how LLM agents have been applied across the DeFi stack. We followed the PRISMA guidelines, searched five scholarly databases, and screened 3,590 records published between 2021 and 2025. Twenty-two peer-reviewed studies met the final inclusion criteria. Thematic synthesis reveals six functional clusters: (i) trading and automated market-making, (ii) portfolio construction and rebalancing, (iii) lending and credit-risk modeling, (iv) governance vote delegation, (v) liquidity-risk assurance and insurance, and (vi) smart-contract security audit. In terms of agent design, single-agent designs still dominate the implementation, but MAS is gaining traction in tasks requiring negotiation or distributed sensing. LLM-powered agents appear mainly in early proof-of-concept security tools.