MarketFish – Simulate a market with 128 AI consumers before you launch

Hacker News Top Tools

Summary

MarketFish is an open-source multi-agent market simulation engine that generates 128+ AI consumers with distinct identities, budgets, and biases to simulate purchase decisions over 30 rounds, helping founders validate product ideas before launch.

No content available
Original Article
View Cached Full Text

Cached at: 07/02/26, 08:04 AM

Key-wxh/market-fish

Source: https://github.com/Key-wxh/market-fish

stars PH v6.0 MIT Python 3.12+ tests 11 providers

🐟 MarketFish

Don't guess. Simulate.

Before you launch, let hundreds of AI consumers vote with their wallets.


MarketFish is a multi-agent market simulation engine. Instead of asking one LLM “will this product succeed?”, it builds a digital market with 128+ AI consumers — each with their own identity, budget, emotions, and biases — and lets them shop across 30 rounds. Their purchase decisions, churn patterns, and social influence reveal what real users would do.

Built on 6 academic papers (Generative Agents, OASIS, TwinMarket, Agent Bazaar, EconSimulacra, SMIF) and 11 LLM providers.

中文文档

Quick Start

git clone https://github.com/Key-wxh/market-fish.git
cd market-fish
cp .env.example .env
# Edit .env — add at least ONE LLM API key (DeepSeek is cheapest)
pip install -r requirements.txt
streamlit run streamlit_app.py

Open http://localhost:8501 → pick a mode → run.

Screenshots

Products Evidence Agents
Agent Graph RL Strategy Coupling

How It Works

Seed Data (static JSON) → 5-Stage Pipeline
  1. Ontology — extract market structure
  2. Knowledge Graph — entities, relationships, pain points
  3. Agent Factory — 128 heterogeneous AI consumers (6 LLMs)
  4. Simulation — 30 rounds: decisions, coupling, RL, memory
  5. Report — evidence: who bought, why, what killed competitors

V6 Modules (6 papers implemented)

ModulePaperWhat it does
MemoryGenerative Agents (UIST 2023)Agents remember purchases, regrets, reflections
Time EngineOASIS (2025)Realistic 24h activation — not all active every round
RecSysOASIS (2025)Personalized product recommendations
BDI v2TwinMarket (NeurIPS 2025)6-step cognitive loop + behavioral biases
StressEconSimulacra (2026)Financial/social pressure → adjusted willingness to pay
GroundingSMIF (ETASR 2026)RAG + rule constraints for realistic decisions

Modes

ModeInputOutput
🔍 ExploreSeed dataAI discovers product directions, ranked
ValidateYour product ideaSurvival score, buyer profiles, optimal price
⚔️ HybridYour product + dataYour idea vs AI competitors, same sandbox

Supported LLM Providers

11 providers. One is enough. More = more diverse agents.

| 🇨🇳 China | DeepSeek, Qwen, Doubao, Zhipu, Baidu, Hunyuan | | 🌍 Global | OpenAI, Anthropic, Google, Mistral, Meta |

CLI

python run.py --mode explore                           # Discover directions
python run.py --mode validate --name "My App" --pricing "$10"  # Test your idea
python run.py --mode explore --reuse-agents            # Reuse agents (save cost)

Project Structure

market-fish/
├── engine/         # Core engine (20+ modules)
│   ├── simulator.py, agent_factory.py    # Simulation core
│   ├── agent_store.py, memory.py         # V6: persistence + memory
│   ├── temporal.py, recsys.py            # V6: time + recommendations
│   ├── bdi_v2.py, stress.py, grounding.py # V6: cognition + stress + validation
├── config/         # Model registry + parameters
├── locales/        # EN/ZH i18n (300+ keys)
├── tests/          # 26/26 tests
├── streamlit_app.py  # Dashboard
├── run.py          # CLI
└── .env.example    # API key template

Academic Foundation

PaperVenueIDModule
Generative AgentsUIST 20232304.03442Memory
OASIS20252411.11581RecSys + TimeEngine
SMIFETASR 202610.48084/etasr.16536Grounding
Agent BazaarPrinceton 20262605.17698RL
TwinMarketNeurIPS 20252502.01506BDI v2
EconSimulacra20262606.26883Stress

vs MiroFish

MiroFish (5.5k ⭐) is the most well-known multi-agent simulation engine. Both projects simulate social/market behavior with AI agents — but with different focuses:

MiroFishMarketFish
ScopeGeneral-purpose social simulationProduct market prediction
ArchitectureFlask + Node.js + DockerStreamlit single-app
MemoryZep Cloud (external service)Built-in (local JSON, zero external deps)
LLMsOpenAI-compatible only11 providers (China + Global)
DataUser-uploaded documents8-source live ingestion pipeline
LanguageEN/ZHEN/ZH
LicenseAGPL-3.0MIT

License

MIT — free for personal and commercial use.


Built by Keystart AI · Solo founder · AI-Native

Similar Articles

666ghj/MiroFish

GitHub Trending (daily)

MiroFish is an open-source swarm intelligence engine that uses multi-agent technology to create a parallel digital world for predicting future outcomes. Users upload seed materials and receive detailed prediction reports and interactive simulations.

Agent Bazaar: Enabling Economic Alignment in Multi-Agent Marketplaces

Hugging Face Daily Papers

Introduces Agent Bazaar, a multi-agent simulation framework for evaluating economic alignment of LLMs, identifying failure modes like algorithmic instability and Sybil deception, and training a 9B model that outperforms frontier models using targeted reinforcement learning.