The hosted MiroFish alternative to self-hosting

Same multi-agent prediction workflow. No GPU, no Docker, no API keys, no server. Run it online in minutes.

MiroFish is an open-source AI prediction engine that simulates thousands of AI agents to forecast how a scenario plays out. Running it yourself means Python, LLM API keys, and a server. MiroFish.us is the hosted alternative — the same seed → knowledge-graph → multi-agent simulation → report workflow, online in your browser, with no setup and pay-per-credit pricing from $2.99.

Self-hosting vs the hosted version

Self-host (open source)MiroFish.us (hosted)
SetupClone repo, Python env, dependencies, configSign in — nothing to install
GPU / Docker / API keysYou provide and manage themNone required
Time to first predictionHours (a real project setup)About 2 minutes
CostVPS + per-token LLM API billing (variable)Credit packs from $2.99 (10 predictions), never expire
MaintenanceKeys, updates, uptime are yoursFully managed
Control & customizationFull — modify the engine, run massive agent countsTuned defaults, run in the browser
PrivacyRuns on your own infrastructurePrivate to your account
Best forDevelopers who want to modify or self-hostAnyone who wants answers, not infrastructure

The open-source engine is excellent — if you want to modify it or self-host, do. This page is for everyone who just wants the prediction. Full breakdown in our guide to using MiroFish online.

Frequently asked questions

Is there a hosted version of MiroFish?

Yes. MiroFish.us runs the full multi-agent prediction workflow — scenario analysis, knowledge graph, simulation, and structured report — as a hosted chat in your browser. There is nothing to install, no GPU or Docker, and no API keys. Credit packs start at $2.99 for 10 predictions.

What is the best MiroFish alternative if I don’t want to self-host?

The hosted version of MiroFish itself. The open-source engine is powerful but requires a Python environment, LLM API keys, and a server. The hosted route gives you the same seed → graph → multi-agent simulation → report workflow with no setup and pay-per-credit pricing.

Does the hosted version cost more than self-hosting?

Usually less, before counting your time. Self-hosting is “free” like a free puppy — you feed the VPS and per-token LLM API bill. Hosted credits start at $2.99 for 10 predictions, typically below the raw API cost of running equivalent simulations yourself.

Can I still self-host if I need full control?

Absolutely — the engine is open source and self-hosting is the right call if you want to modify it, run massive agent counts, or keep everything on your own metal. The hosted version is for people who want the answers without the infrastructure.

Skip the setup — predict now

No install, no API keys. Credit packs from $2.99, credits never expire.

Start predicting