MiroFish
mirofish.workMulti-agent simulation for scenario prediction
Turn source material into inspectable prediction reports with multi-agent simulation.
Overview
MiroFish is an AI prediction engine that turns source material into a world model, runs simulated agents, and returns an inspectable prediction report. It is designed for teams and individuals facing uncertain decisions, such as launches, policy moves, market narratives, crisis responses, or story-world outcomes. The platform supports text, PDF, Markdown, TXT, and decision notes as input, and outputs prediction reports, branch points, and follow-up chat.
How it works
MiroFish follows a five-step workflow: Graph Building extracts seed facts into a GraphRAG knowledge graph, Environment Setup creates relationships and agent personas, Simulation runs agents through interaction rounds with temporal memory updates, Report Generation turns the post-simulation state into structured findings, and Deep Interaction lets users chat with simulated agents or the ReportAgent. The platform supports natural language input and can handle text, PDF, Markdown, TXT, and decision notes as source material.
Key capabilities
Key capabilities include MiroFish Swarm Intelligence, where independent AI agents interact in a simulated social environment, and GraphRAG, which extracts entity relationships from source material. Users can inject policy changes or market shocks mid-simulation through God's-Eye Variable Injection, and each run produces structured prediction reports with narrative analysis, group dynamics, probabilities, timeline notes, and follow-up paths. The platform also supports deep agent interaction for inspecting motives and evidence.
Who it is for
MiroFish is for teams and individuals who need to rehearse decisions under uncertainty. It is used for public opinion and crisis management, financial scenario forecasting, policy impact testing, and creative narrative exploration. The platform is suitable for PR and communications, finance and investing, government and policy, and writers and creators, helping them simulate how news, crises, or announcements spread and where sentiment may shift.
Pricing and limits
MiroFish offers a free trial with 1 guest simulation or 2 simulations for signed-in accounts, using DeepSeek V4 Flash only. Paid plans include Starter at $9/month with 10 analyses and up to 50 agents, Pro at $29/month with 50 analyses and up to 100 agents, and Enterprise at $59/month with 320 analyses and up to 200 agents. Yearly billing is available at 50% off, and the platform is open source under AGPL-3.0 with Docker options for self-hosting.
Key features
Multi-agent simulation
Independent AI agents interact in a simulated social environment to reveal dynamics a single model can miss.
GraphRAG knowledge graphs
Extract entity relationships from source material into a graph that anchors agent memory and reasoning.
Natural language interface
Describe the scenario in plain language; MiroFish handles setup, simulation, and reports in one conversation.
Structured prediction reports
Each run produces narrative analysis, group dynamics, probabilities, timeline notes, and follow-up paths.
God's-Eye variable injection
Inject policy changes, market shocks, or new assumptions and watch the simulated world recalibrate.
Deep agent interaction
After simulation, interview agents or query the ReportAgent to inspect motives and evidence.
Use cases
Public opinion and crisis management
Simulate how news, crises, or announcements spread and where sentiment may shift.
Financial scenario forecasting
Model sentiment after earnings, regulation, or macro signals.
Policy impact testing
Test policy drafts and compare which groups may support, resist, or reshape debate.
Creative narrative exploration
Explore story branches or fictional societies driven by character behavior.
Scenario planning
Plan launches, pricing changes, and operating decisions by comparing base, upside, downside, and surprise branches.
Pricing & limits
From the official site · checked 10/10/2026Frequently asked questions about MiroFish
- What is MiroFish?
- MiroFish is an AI prediction engine that turns source material into a world model, runs simulated agents, and returns an inspectable prediction report.
- Who is MiroFish for?
- It is for teams and individuals who need to test scenarios and predict outcomes in areas like public opinion, market narratives, policy debates, launches, and creative story worlds.
- Is MiroFish free to use?
- Yes. Guest visitors get 1 free simulation, and signed-in free accounts get 2 free simulations using DeepSeek V4 Flash only.
- What platforms does MiroFish support?
- MiroFish is a web-based platform. It is also open source under AGPL-3.0 and can be self-hosted with Docker.
- How does MiroFish simulate predictions?
- MiroFish follows five steps: Graph Building, Environment Setup, Multi-agent Simulation, Report Generation, and Deep Interaction with simulated agents or the report.
Related tools

ProbSee
Find prediction market opportunities and pressure-test your views with ProbSee.

ConceptFrameworkAI
Describe your research in plain words and get a publication-ready conceptual framework diagram in seconds.

paper banana
Turn paper text into research figures and scientific diagrams with reference-guided planning and iterative refinement.

Blix
Turn open-ended survey responses and reviews into coded data and insights in minutes.

chartcastr
Connect your SaaS tools and spreadsheets, get AI analysis and next steps delivered to Slack.

TrafficClaw
Ask plain-English questions about your Google Analytics and Search Console data and get instant answers.

Football AI Prediction King
Compare free football AI predictions with likely scores, win/draw/loss picks, and model consensus for daily matches.

BizIntelligenceChampions
Turn natural-language prompts into working, interactive charts in Power BI and Excel.