Deepmiro
DeepMiro runs multi-agent simulations to predict how communities respond to events or policy changes. It extracts entities from documents, spins up hundreds of AI agents with distinct personalities and social networks, then simulates interaction across Twitter-like and Reddit-like platforms. The MCP exposes tools to trigger predictions, query simulation status, retrieve structured reports, and chat with individual agents post-simulation. Behind the scenes it uses GraphRAG for entity extraction, TWHIN-BERT for social embeddings, and SurrealDB for agent memory and relationships. Useful when you need emergent behavior forecasting that goes beyond single-prompt analysis, like testing policy drafts, market scenarios, or narrative outcomes before they play out in the real world.
- GitHub stars
- 7
- tools
- 9
- auth
- Secret needed
- transport
- STDIO
- env vars
- 2
- last updated
- Apr 9, 2026
Tools
9 toolsPublic tool metadata for what this MCP can expose to an agent.
create_simulationRun a full swarm prediction. Builds a knowledge graph, generates agent personas, runs a multi-agent social media simulation, and generates a prediction report. Streams progress updates. Returns the final report when complete.6 params
Run a full swarm prediction. Builds a knowledge graph, generates agent personas, runs a multi-agent social media simulation, and generates a prediction report. Streams progress updates. Returns the final report when complete.
agent_countintegerdocument_idstringplatformstringtwitter · reddit · bothpresetstringquick · standard · deeppromptstringroundsintegersimulation_statusCheck the progress of a running or completed simulation. Returns phase-aware status with entity names and action content. Phases: building_graph → generating_profiles → simulating → completed.2 params
Check the progress of a running or completed simulation. Returns phase-aware status with entity names and action content. Phases: building_graph → generating_profiles → simulating → completed.
detailedbooleansimulation_idstringget_reportGenerate and retrieve the prediction report for a completed simulation. If the report hasn't been generated yet, triggers generation (may take 1-3 minutes). Returns a detailed markdown analysis of the simulation results.1 params
Generate and retrieve the prediction report for a completed simulation. If the report hasn't been generated yet, triggers generation (may take 1-3 minutes). Returns a detailed markdown analysis of the simulation results.
simulation_idstringinterview_agentChat with a specific simulated agent to understand their perspective, reasoning, and predicted behavior. The agent responds in character based on their persona and simulation experience.4 params
Chat with a specific simulated agent to understand their perspective, reasoning, and predicted behavior. The agent responds in character based on their persona and simulation experience.
agent_idintegermessagestringplatformstringtwitter · redditsimulation_idstringlist_simulationsList past simulation runs with their status and metadata.1 params
List past simulation runs with their status and metadata.
limitintegersearch_simulationsSearch past simulations by topic, project name, or simulation ID.1 params
Search past simulations by topic, project name, or simulation ID.
querystringquick_predictFast, lightweight prediction without running a full simulation. Uses the LLM to simulate swarm behavior and predict outcomes. Returns in seconds. For deeper analysis, use create_simulation instead.1 params
Fast, lightweight prediction without running a full simulation. Uses the LLM to simulate swarm behavior and predict outcomes. Returns in seconds. For deeper analysis, use create_simulation instead.
promptstringupload_documentUpload a document for use in simulations. LIMITS: Max 10MB, PDF/MD/TXT only. The server extracts text server-side (PyMuPDF for PDFs). Returns a document_id to pass to create_simulation. NOTE: Only works with local file paths (stdio transport). For remote/hosted mode, the clien...1 params
Upload a document for use in simulations. LIMITS: Max 10MB, PDF/MD/TXT only. The server extracts text server-side (PyMuPDF for PDFs). Returns a document_id to pass to create_simulation. NOTE: Only works with local file paths (stdio transport). For remote/hosted mode, the clien...
file_pathstringsimulation_dataAccess raw simulation data: agent profiles, configuration, action logs, social media posts, round-by-round timeline, per-agent activity stats, and interview history. Use this to inspect what happened during a simulation.6 params
Access raw simulation data: agent profiles, configuration, action logs, social media posts, round-by-round timeline, per-agent activity stats, and interview history. Use this to inspect what happened during a simulation.
action_typestringagent_namestringdata_typestringprofiles · config · actions · posts · timeline · agent_statslimitintegerplatformstringtwitter · redditsimulation_idstringSponsors
Agent CaffeineSponsoredClose your lid, come back to finished work
SnugSponsoredThe clipboard manager for the AI eraPromote your product hereReach 125k+ visitors monthly
WaxSponsoredBlock anything with supervision, without erasing your iPhonePromote your product hereReach 125k+ visitors monthly
knobsSponsoredSee your site on any phone, any day, in any languageConfiguration
DEEPMIRO_API_KEYsecretYour DeepMiro API key (get one free at https://deepmiro.org)
MIROFISH_URLOverride for self-hosted engine URL (default: https://api.deepmiro.org)
Readme
A swarm intelligence engine that rehearses the future.
Feed it a document. Describe a scenario. Watch hundreds of AI agents with distinct personalities, memories, and social instincts interact — and return with a prediction.
One-click self-host — four services, one API key, ~60 seconds. Full walkthrough ↓
Contents
Details
- Repository
- github.com/kakarot-dev/deepmiro
- Package
deepmiro-mcp- Transport
- STDIO
- Auth
- Required
- Registry
- MCP registry, active