Deepmiro

Lets an assistant run large-scale multi-agent simulations that predict how communities react to events and policies.

What you can do

Deepmiro is a local MCP server that simulates hundreds of AI agents to forecast community reactions to events and policies. It runs as an npm package over stdio and talks to a hosted simulation engine, with an option to point at a self-hosted engine instead. The manifest does not list individual tools, so the exact operations exposed to the assistant are not documented here.

When to use it

Worth adding when you want an assistant to explore how a population or community might respond to a proposed event, announcement, or policy before committing to it. It suits scenario and sentiment exploration rather than factual lookup.

Requirements

Runs locally as the npm package deepmiro-mcp over stdio, so a Node.js runtime is needed. Requires a DeepMiro API key supplied through the DEEPMIRO_API_KEY environment variable. Network access to the default engine endpoint is required unless the MIROFISH_URL variable is set to a self-hosted engine.