ChatArena (or Chat Arena) is a Multi-Agent Language Game Environments for LLMs. The goal is to develop communication and collaboration capabilities of AIs.

🚀 Introducing 🐪 CAMEL: a game-changing role-playing approach for LLMs and auto-agents like BabyAGI & AutoGPT! Watch two agents 🤝 collaborate and solve tasks together, unlocking endless possibilities in #ConversationalAI, 🎮 gaming, 📚 education, and more! 🔥

Here we will keep track of the latest AI Game Development Tools, including LLM, World Model, Agent, Code, Image, Texture, Shader, 3D Model, Animation, Video, Audio, Music, Singing Voice and Analytics. 🔥

[ICLR 2026] LLM/VLM gaming agents and model evaluation through games.

CivAgent is an LLM-based Human-like Agent acting as a Digital Player within the Strategy Game Unciv.

Implementation of the board game Codenames, re-imagined as a collaborative game between LLM agents

How to create rational LLM-based agents? Using game-theoretic workflows!

An LLM agent built on top of OpenAI that predicts winners for ESPN's Pick-Em's game.

This is a repo for studying the application of LLM Agents on Games

LieGraph - AI Agent-Driven "Who Is Spy" Game. Multi-agent social deduction game built with LangGraph where autonomous AI agents use LLM reasoning to find the spy through conversation analysis and strategic voting.

Public Goods Game (PGG) Benchmark: Contribute & Punish is a multi-agent benchmark that tests cooperative and self-interested strategies among Large Language Models (LLMs) in a resource-sharing economic scenario. Our experiment extends the classic PGG with a punishment phase, allowing players to penalize free-riders or retaliate against others.

Build RL+LLM multi-agent simulations in game engines for realistic human behaviors and studying cities

Make open-weight LLM agents play the game "Among Us", and study how the models learn and express lying and deception in the game.

AI Sandbox Game: A highly customizable, local-first text RPG driven by LLMs. Create world cards, configure multi-step AI agents, and play with zero backend.

A method to use LLM function calling with real-time level speed, empowering latency-sensitive applications like robotics, game AI, and autonomous agents.

This is a repository containing the implementation of the One Night Ultimate Werewolf game environment, and an RL-instructed LLM-based agent framework.

Pure LLM agent powered card game w/ TTS and live dashboard.

AmongAgents, a text-based game environment for LLM agent evaluation inspired by the popular game Among Us

Capabench：A Game-Theoretic Evaluation Benchmark for Modular Attribution in LLM Agents

implements the methodology outlined in the paper *Cultural Evolution of Cooperation among LLM Agents*. The paper explores whether a society of large language model (LLM) agents can develop cooperative norms through cultural evolution, using the classic *Donor Game*. The goal is to evaluate multi-agent interaction dynamics

AI agent that plays The Battle of Polytopia using LLM-powered strategic reasoning. Training environment for AI strategy game research.

Generative agents — computational software agents that simulate believable human behavior and OpenAI LLM models. Our main focus was to develop a game - “Werewolves of Miller’s Hollow”, aiming to replicate human-like behavior.

A sample Unity project where I integrated the ability to chat with agents powered by an LLM, capable of choosing which predefined actions to perform in the game.

sts2-agent is an agent/mod prototype for Slay the Spire 2. The repository focuses on exposing structured in-game state, safe action execution, and local tooling so external policies or LLMs can drive live gameplay.

A sims game/architecture to simulate multi LLM powered autonomous agents.

In this repo I am exploring game theory with LLM agents. Have Fun.

Code for "TowerMind: A Tower Defence Game Learning Environment and Benchmark for LLM as Agents (AAAI 2026 Oral)"

DSGBench is a game benchmark designed to evaluate LLM agents across diverse, dynamic environments, including games like StarCraft II and Werewolf. It tests agents' abilities in decision-making, strategy, adaptability, and interaction, offering detailed performance metrics to improve LLM optimization in complex tasks.

A multi-agent benchmark where eight LLMs play a money-driven elimination game with private transfers and a buyout endgame, and are ranked by final wealth

Taboo game made with AutoGen using OpenAI LLM agents.