Concordia is an open-source library from Google DeepMind for building generative agent-based simulations. It has 1,810 stars on GitHub and is under active development in 2026. Concordia lets you populate a simulated physical, social, or digital environment with LLM-powered agents and watch how they interact over time.
Concordia borrows its interaction pattern from tabletop role-playing games. A special entity called the Game Master simulates the environment: player entities describe their intended actions in natural language, and the Game Master translates those intentions into plausible outcomes, checking consistency with the simulated world. Agents are assembled from components that supply memory, reasoning, and behavior, and the library connects to standard LLM APIs. DeepMind documents the design in two technical reports and positions Concordia for research in the social sciences and AI evaluation. Simulations are configured in Python: you define the entities, their components, and the Game Master’s responsibilities, then step the world forward while agents perceive, deliberate, and act. Every run produces full interaction logs, which makes experiments auditable and repeatable.
Concordia is free and open source under the Apache-2.0 license, installable from PyPI as gdm-concordia. The only cost is the LLM API usage your simulations consume.
Concordia is for researchers in computational social science, AI safety, economics, and cognitive science who want to study how LLM agents behave in shared environments, and for teams evaluating services through simulated users.