SimpleAIsimpleai
mirofish

MiroFish

Free tierUpdated 2026-04

Simulate the future with thousands of AI agents acting out real-world scenarios.

🔴Advanced1 hour to set upTry MiroFish

What is it?

MiroFish is an open-source AI simulation platform that lets you build digital worlds populated by thousands of intelligent agents — each with their own personality, memory, and decision-making logic. These agents interact with each other autonomously to act out complex real-world scenarios.

The idea is simple: instead of guessing what might happen, you simulate it first. Feed MiroFish a scenario — a business decision, a policy change, a market event — and watch how thousands of AI-powered "people" respond to it. The results can surface outcomes you hadn't considered.

It uses a technology called GraphRAG to map relationships between entities (people, companies, events) and runs parallel simulations so you can compare multiple outcomes side by side.

Who is it for?

  • Policy researchers and analysts who want to model how decisions ripple through systems
  • Business strategists stress-testing plans before committing
  • Financial analysts exploring market behaviour under different conditions
  • Writers and game designers building worlds with believable, autonomous characters
  • Academics and journalists investigating complex social or political dynamics

The magic moment

You define a scenario: "A major tech company announces layoffs of 10,000 people." MiroFish spins up thousands of agents — employees, investors, journalists, competitors, customers — and simulates how each one reacts, how the news spreads, and what second-order effects emerge. The ReportAgent then summarises what happened and answers your follow-up questions. It's like running a dress rehearsal for reality.

Step-by-step setup

  1. Install Node.js 18+ and Python 3.11 or 3.12 on your machine
  2. Clone the MiroFish repo from GitHub
  3. Install the UV package manager
  4. Add your LLM API key (Claude or any OpenAI-compatible model)
  5. Set up Zep Cloud for agent memory (free tier available)
  6. Run npm run dev to launch the platform
  7. Define your scenario and start your first simulation

This is an advanced setup — budget around an hour and expect some troubleshooting.

Compare with similar tools

  • NotebookLM — great for analysing documents you give it; MiroFish goes further by simulating how agents act on information dynamically
  • Perplexity — searches the real web for answers; MiroFish simulates hypothetical futures that don't exist yet
  • Obsidian — a knowledge base you build over time; MiroFish is an active simulation engine, not a note-taking tool