Multi-agent AI research
Research infrastructure
for interacting AI agents.
Build, simulate, and study populations of intelligent agents—and the collective behavior that emerges between them.
THE PLATFORM
AgentsNet is a research and engineering platform for building and studying systems of interacting AI agents.
Give agents memory, beliefs, and the ability to act within a shared world.
Explore interactions across environments, institutions, and networks.
Turn collective behavior into questions we can observe, test, and understand.
01 / PROJECTS
A growing collection of simulation frameworks, research tools, and experimental systems.
An open-source framework for LLM-powered multi-agent simulation.
Agent / environment loop
Individual decisions. A shared environment.
A startup simulation with 80,000 agents. Watch company growth, recruiting, fundraising, and evolving networks in a recorded replay.
Founders / company network
80,000 agents. One evolving startup economy.
Explore live markets alongside a simulated market of trading agents. Follow prices, news, and the dynamics of investor decisions.
Investors / market dynamics
Market signals. Collective decisions.
A future direction for studying agent behavior and evaluating multi-agent systems.
A future direction for making agent interactions and network dynamics easier to explore.
AgentSims is the first project in the AgentsNet ecosystem.
02 / RESEARCH
Six research directions guide how we build, study, and evaluate multi-agent systems.
Study how populations of autonomous agents interact and evolve in shared environments.
Explore how memory, beliefs, and internal state shape decisions and behavior.
Analyze coordination, diffusion, polarization, and the dynamics of collective behavior.
Build infrastructure for reproducible computational experiments with AI agents.
Study how people and agents influence each other in social, economic, and political environments.
Measure behavior, reliability, alignment, and emergent failure modes across agent systems.
03 / BUILT IN THE OPEN
We release core research infrastructure and selected projects as open source, so others can inspect, extend, and build on the work.
Explore AgentSimsExplore the project for documentation and available releases.
04 / ABOUT AGENTSNET
AgentsNet explores AI as populations of interacting agents embedded in environments, institutions, and networks.
We build open research infrastructure for studying these systems—and the behavior that emerges between them.
Get access to latest updates, contact Fan at fhuang181@gmail.com.
Agents in context