MiroFish

MiroFish is developed by GitHub user 666ghj and the MiroFish team, and the README says it is supported and incubated by Shanda. It calls itself "a next-generation AI prediction engine powered by multi-agent technology." You upload some seed material, such as news, a report or a scenario, and it builds a relationship graph from it, then creates a group of AI agents with memory and personas and lets them interact in a simulated environment. When the run ends you get a prediction report and an interactive sandbox you can keep questioning and adjusting. The README mentions "financial signals," and its financial prediction examples are marked "coming soon." The documentation is in Chinese and English.

Who it's for:developers and researchers who want to try scenario simulations with many agents and are willing to set up the environment themselves. Not useful if you don't write code. We don't recommend using it for investment decisions.

Before you start: you need Node 18 or newer, Python 3.11 to 3.12, uv, an LLM API key and a Zep Cloud account; start it with npm run setup:all and npm run dev, or use Docker. Uploaded material is sent to an external model service (the README recommends Qwen-plus) and to Zep Cloud. Runs with many simulation rounds use up API credit quickly. Licensed under AGPL-3.0. Based on the project's public documentation. Not tested by AI Inspiration World.