AI Geopolitical Wargaming & Policy Simulation Projects
April 20, 2026 · View on GitHub
A curated list of projects using AI, LLMs, and multi-agent systems for geopolitical wargaming, policy simulation, conflict modeling, and strategic forecasting.
Last updated: 2026-04-19
See also: AI-Synthesised-Perspectives for multi-agent deliberation and debate tools, and AI-Decision-Eval-Projects for decision-making evaluation frameworks.
Table of Contents
- Wargaming & Conflict Simulation
- Policy Simulation
- Multi-Agent Geopolitical Systems
- Strategic Forecasting
- International Relations Modeling
- Frameworks & Tools
- Reference Data & Codebooks
Wargaming & Conflict Simulation
Projects focused on AI-driven military and geopolitical wargaming scenarios.
WargamesAI
Professional Wargaming LLM Toolbox
Language: Python
Author: user1342
Snowglobe
Open-ended wargames with large language models. Multi-agent system where every stage from scenario preparation to post-game analysis can be carried out by AI, humans, or a combination.
Language: Python
Author: IQTLabs (In-Q-Tel)
Geopol Modeller
Fork of IQTLabs Snow Globe, remodeled around OpenRouter as a single LLM gateway with model pools, a Tavily-powered planning agent, LangGraph simulation loop, Rich progress UI, checkpointing, Typst PDF reports, and edge-TTS podcast generation. Stripped LangChain/torch/transformers in favor of direct OpenAI SDK calls.
Derived from IQTLabs/snowglobe.
Language: Python
Author: danielrosehill
ROMANCER
RAND Ontological Model for Assessing Nuclear Crisis Escalation Risk. Agent-based simulation with multiple theories-of-mind for exploring how decisionmaker actions, threats, and demands affect nuclear escalation outcomes.
Language: Python
Author: RANDCorporation
LLMWargaming
LLMs for Wargames. Experiment comparing human players to LLM-simulated responses in a fictional US-China crisis escalation scenario with 107 national security experts.
Language: Julia
Author: ancorso
LLM-CBRN-Risks
[ACL 2025 Findings] Nuclear Deployed: Analyzing Catastrophic Risks in Decision-making of Autonomous LLM Agents. Studies how LLM agents behave in nuclear crisis scenarios.
Language: Python
Author: pillowsofwind
TIDE2024 LLM Wargame
Wargame simulations using LLMs. Created during TIDE Hackathon 2024. Customizable scenarios with support for simulations across all five military domains.
Language: Python
Author: MilaSong
Policy Simulation
Tools and frameworks for simulating policy outcomes using AI.
ThunderBeans
AI-driven parliamentary simulation + SendGrid-powered report delivery. Built with ChatGPT, Replit, Node.js, and policy simulation logic.
Language: JavaScript
Author: TawandaVera
Global Microeconomic Simulation Engine
OmniEcon Nexus is an open-source, high-performance simulation engine for global micro/macro-economic analysis. Using deep learning, agent-based modeling, and optimization, it supports 5M agents for forecasting, risk analysis, policy simulation, and portfolio management.
Language: Python
Author: vinhatson
Election Agent
LLM-powered agentic system for Israeli Knesset election analysis -- comparing 4 tool routing strategies (DS-UA 301)
Language: Python
Author: YardenMorad2003
Multi-Agent Geopolitical Systems
Multi-agent AI systems simulating geopolitical actors and interactions at scale.
OASIS
Open Agent Social Interaction Simulations with One Million Agents
Language: Python
Author: camel-ai
MiroFish
A Simple and Universal Swarm Intelligence Engine, Predicting Anything
Language: Python
Author: 666ghj
MiroFish-Offline
Fully local, English-language fork of MiroFish. Multi-agent swarm intelligence engine that simulates public opinion, market sentiment, and social dynamics entirely on local hardware with zero cloud dependencies. Uses Ollama + Neo4j Community Edition.
Community English fork of the original Chinese MiroFish project, making it accessible to a broader audience and fully self-hostable.
Language: Python
Author: nikmcfly
MiroShark
Universal Swarm Intelligence Engine. Multi-agent simulation engine that generates hundreds of AI agents to simulate public reaction across Twitter, Reddit, and Polymarket simultaneously. Features cross-platform simulation with market-media feedback bridge.
Evolution of the MiroFish concept adding Polymarket prediction market trading alongside social media simulation.
Language: Python
Author: aaronjmars
Hermes Geopolitical Market Sim
Hermes skill package for geopolitical market forecasting. Wires together WorldOSINT headless feeds, Polymarket pricing, MiroFish simulation, and Hermes operator workflows. Features counterfactual branching to inject new actors into past simulations and measure butterfly effects.
Orchestration layer combining OSINT feeds, prediction markets, and multi-agent simulation into one pipeline.
Language: Python
Author: nativ3ai
AI Diplomacy
Frontier Models playing the board game Diplomacy. Each power controlled by an autonomous LLM agent that maintains state, forms relationships, conducts negotiations, and makes strategic decisions.
Uses the board game Diplomacy as substrate for real diplomatic negotiation, alliance formation, and strategic planning — core skills for geopolitical simulation.
Language: Python
Author: GoodStartLabs
AgentSociety
Large-scale social simulation (10k+ agents, 5M interactions) to understand human behaviors and society through LLM-driven agents. Agents have emotions, needs, motivations, and perform social behaviors including economic activity.
From Tsinghua University. Can model population-level responses to policy changes, economic shocks, or political events.
Language: Python
Author: tsinghua-fib-lab
Political Actor Agent (PAA)
[AAAI 2025] Simulating Legislative System for Roll Call Votes Prediction with Large Language Models. Constructs LLM agents representing legislators with profiles, constituency data, and voting history.
Directly simulates a legislative system with LLM-powered political actors, modeling party influence and roll-call voting.
Language: Python
Author: Lihaogx
LLM Agent Opinion Dynamics
Simulating Opinion Dynamics with Networks of LLM-based Agents. Models how opinions spread and shift through social networks of AI agents.
Relevant to understanding information warfare, political polarization, and mobilization dynamics.
Language: Python
Author: yunshiuan
Strategic Forecasting
AI systems for geopolitical forecasting and scenario analysis.
Geopol Forecaster
Two-stage geopolitical forecasting pipeline combining Snowglobe-style actor simulation (38 actors, 4 timesteps) with LLM Council 6-lens analytical panel. Seeded with live Tavily/ISW news data. Produces calibrated probabilistic forecasts with credible intervals.
Language: Python
Author: danielrosehill
Geopol Forecast Council
Lean spin-off of Geopol Forecaster that skips the actor simulation. Uses an explicit grounding tool list (RSS + Perplexity Sonar + Tavily) to build a timestamped SITREP, then polls a five-model panel (GLM, DeepSeek, Gemini, Claude, Kimi) for three concrete predictions per horizon with supporting/counterveiling historical precedents, change factors, and confidence values. Structured-output JSON schema, per-stage checkpointing, deterministic Typst rendering with cover page, confidence bars, tone classification, and absolute UTC horizon timestamps. Main + full-verbatim PDF variants.
Cheap cousin of Geopol Forecaster — trades actor-level emergent dynamics for short-horizon news-grounded panel forecasts.
Language: Python
Author: danielrosehill
Geopol Forecasts Index
Index of prediction runs from Geopol Forecaster with accuracy tracking. Links to open-sourced reports including the Iran-Israel-US ceasefire durability assessment (09/04/2026).
Author: danielrosehill
Iran-Israel-Ceasefire-Prediction-090426
Open-sourced forecast run assessing the April 8 2026 Iran-Israel-US ceasefire. Notable for 6-point convergence between independent simulation (28%) and council (22%) stages at the 72-hour mark. Includes full report, code snapshot, and data visualizations.
Author: danielrosehill
Geopol Forecaster POC
Experimentary prediction analysis for real world events (Iran Israel)
Language: TypeScript
Author: danielrosehill
Geopolitical And OSINT Index
Index of repositories related to OSINT, geopolitical analysis, and situational awareness
Author: danielrosehill
Israel GeoInsights
Experimentary repository; using LLMs to analyse geopolitical developments in Israel & the Middle East
Language: Python
Author: danielrosehill
International Relations Modeling
Projects modeling international relations dynamics with AI agents.
Claude Georeaction Researcher
Claude template for analysing global sentiment/reaction to a geopolitical issue
Language: Python
Author: danielrosehill
Frameworks & Tools
General-purpose frameworks and tools applicable to geopolitical AI simulation.
Geopolitical Analysis System Prompts
System prompts for geopolitical related AI assistants and tools
Author: danielrosehill
Iran-Israel-Grounding-Package
Curated whitelist of URLs and sources for AI agents grounding analysis of the Iran–Israel–US conflict (2026). Includes a SITREP-generation skill and a LangGraph pipeline design (3-hour triage → deep-research escalation → SITREP → analyst) intended as the grounding layer for Geopol Forecaster-style forecast producers.
Author: danielrosehill
OpenPolicyStack
Open-source modular stack for AI-enhanced policy making. Includes reusable microservices for data ingestion, simulation, analytics, dashboards, strategy agents, and decision-support tools.
Language: Python
Author: OscarDiez
Agent Council
Multi-agent collaboration plugin for Claude Code -- orchestrate multiple AI agents (Codex CLI, Gemini CLI, etc.) for diverse perspectives
Language: JavaScript
Author: team-attention
Brainstormers
A suite of specialized agents designed to power brainstorming sessions. Each agent brings a unique perspective and tailored approach to generate creative, targeted solutions for diverse challenges.
Language: TypeScript
Author: Azzedde
AI Agent Virtual Panel Configs
Sets of 'panels' for testing Rally (AskRally.com) exploring various 'frontier' applications for the concept of AI personas and simulated voting bodies and thinking groups
Language: Shell
Author: danielrosehill
Reference Data & Codebooks
Datasets, codebooks, and reference schemas used to ground geopolitical AI projects.
CAMEO Codebook
Unofficial machine-readable (JSON/YAML) extraction of the CAMEO (Conflict and Mediation Event Observations) codebook used by GDELT, ACLED, and other political event data projects. Consolidates the 1.1b3 PDF manual with the full GDELT lookup set (310 event codes with Goldstein scale, 261 country codes, 646 ethnic groups, 117 known international orgs, 31 religion codes, 40 actor-type codes) into a single unified schema.
Useful as a grounding reference for any LLM-based geopolitical event coding, actor classification, or conflict-data pipeline.
Author: danielrosehill
World-Geopol-Fora
Machine-readable data about real-world geopolitical fora (UNSC, UNGA, NATO, EU Council, G7/G20, BRICS, SCO, OAS, AU, Arab League, OIC) — membership, mandates, voting rules, and decision dynamics — intended as ground-truth context for multi-agent AI simulations modelling political reactions. UN bodies seeded from AI-Agent-UN; other fora stubbed against a shared JSON Schema.
Author: danielrosehill
Contributing
Anyone is welcome to open a pull request to add a project to this list. Drop me a line at public@danielrosehill.com if you'd like me to add it manually.
Disclaimer
This resource is intended for those discovering AI-powered geopolitical simulation and wargaming projects. It is not exhaustive and is maintained on a best-effort basis. The inclusion of a project does not constitute an endorsement.
Maintained by Daniel Rosehill