AgentVerse is an open-source framework from OpenBMB for deploying multiple large language model agents that collaborate on tasks or inhabit simulated worlds. Published on GitHub, where it has gathered more than 5,100 stars, it provides two complementary frameworks, one for task-solving agent teams and one for agent simulation, under the Apache-2.0 license.
Single agents hit limits quickly on complex work. AgentVerse explores what happens when many LLM-based agents with different roles cooperate: in its task-solving framework, agents form groups that discuss, delegate and iteratively refine solutions for jobs such as software development, consulting-style problem solving and creative writing. Its simulation framework instead places agents in shared environments, from classroom and prisoner’s-dilemma style social experiments to game worlds, so researchers can observe emergent multi-agent behavior. The framework abstracts agent roles, environments, communication and memory, and works with GPT-class models through configurable backends.
AgentVerse has been used in a line of research papers on multi-agent collaboration, and its modular environment design makes it straightforward to add new tasks, new agent types and new communication protocols. Sample configurations for group problem solving and social experiments are included, so a working multi-agent scenario can be running within minutes of installation.
AgentVerse is completely free and open source under the Apache-2.0 license. You only pay for the model API calls your agents make; the framework itself has no paid tier or hosted plan.
AgentVerse is for researchers and developers experimenting with multi-agent systems: teams prototyping collaborative agent workflows, and academics studying how groups of language-model agents solve problems or interact socially.