Predict Any Outcome

A multi-agent simulation engine that turns your scenarios into structured predictions — powered by the LLMs you already use.

Launch App

What LeanSwarm Does

🔒

Bring Your Own Keys

Use your existing OpenAI, Anthropic, or any OpenAI-compatible API keys. Nothing is stored on our servers.

👥

Multi-Agent Simulation

Define personas with perspectives and biases. Agents debate, reason, and converge on outcomes in realistic world models.

📈

Structured Reports

Get probability estimates, confidence intervals, key factor analysis, and divergent viewpoints — all in one report.

🔎

Post-Simulation Explorer

Drill into every agent's reasoning chain. See who changed their mind, why, and what evidence swayed the outcome.

How It Works

1

Describe Your Scenario

Write a question and define the world model — stakeholders, constraints, and what "success" means.

2

Agents Simulate

Each agent adopts a persona, researches the scenario, debates others, and iteratively refines its prediction.

3

Get the Prediction

Review the aggregate forecast, drill into agent reasoning, and export the structured report.