Abstract
Recent advances on large language models (LLMs) enable researchers anddevelopers to build autonomous language agents that can automatically solvevarious tasks and interact with environments, humans, and other agents usingnatural language interfaces. We consider language agents as a promisingdirection towards artificial general intelligence and release Agents, anopen-source library with the goal of opening up these advances to a widernon-specialist audience. Agents is carefully engineered to support importantfeatures including planning, memory, tool usage, multi-agent communication, andfine-grained symbolic control. Agents is user-friendly as it enablesnon-specialists to build, customize, test, tune, and deploy state-of-the-artautonomous language agents without much coding. The library is alsoresearch-friendly as its modularized design makes it easily extensible forresearchers. Agents is available at https://github.com/aiwaves-cn/agents.