Title: 分享一个 AI 多智能体场景模拟工具： MiroFish | BestBlogs.dev
URL Source: https://www.bestblogs.dev/article/1cfbb691?amp%3Butm_medium=feed&%3Butm_campaign=resources&%3Bentry=rss_article_item
Published Time: 2026-07-08 15:14:44
Markdown Content: Skip to main contentAudio 2 
Search Ctrl+K
Change language Switch ThemeSign In
[](https://www.bestblogs.dev/en/explore/brief "Daily Brief")[](https://www.bestblogs.dev/en/explore/newsletter "Weekly Picks")[](https://www.bestblogs.dev/en/explore/topics "Topics")[](https://www.bestblogs.dev/en/worldcup "World Cup Special")
[](https://www.bestblogs.dev/en/settings "Settings")[](https://www.bestblogs.dev/en/docs "Help Center")
Narrow Mode
75
分享一个 AI 多智能体场景模拟工具： MiroFish
This article introduces an AI multi-agent scenario simulation tool, MiroFish, which uses GraphRAG knowledge graphs to construct agents with different stances for interactive simulation, outputting structured prediction reports suitable for sandbox rehearsals before product, policy, and market decisions. V V2EXFollow·
Today·714 words (about 3 min)
AI Summary & Key Points
Summary
The author shares the AI scenario prediction tool MiroFish they developed. Unlike traditional Q&A-style AI, this tool converts user-described scenarios in natural language (which can be supplemented with PDF, Markdown, etc.) into a GraphRAG knowledge graph, generating AI agents with different stances, memory, and behavioral logic that interact in a simulated environment, ultimately outputting a structured report containing possible paths, key assumptions, risk points, timelines, opposing views, and validation recommendations. The author believes its value lies in observing disagreements and group reactions through multi-agent interaction, supporting follow-up questioning and comparison after assumption modifications, and providing both open-source and hosted usage options. It is applicable to scenarios such as product launches, price increases, brand sentiment, policy discussions, and investment research.
Main Points
* 1. MiroFish performs predictions through multi-agent interaction, rather than a single model directly drawing conclusions.
The system constructs agents with different stances that interact in a simulated environment, observing disagreements, propagation paths, and group reactions, and outputs a structured report.
* 2. Supports multiple input formats and subsequent interaction.
Accepts natural language, PDF, Markdown, etc. as seed material, and supports follow-up questioning and re-running comparisons after assumption modifications.
* 3. Suitable for pre-decision sandbox simulations.
Suitable for scenarios involving public reaction and interest conflicts, such as product, policy, and market contexts, e.g., analyzing user reactions to a 20% price increase.
Sign in to highlight text and take notes as you read. Sign in now
最近整理了一个 AI 场景预测 / 多智能体模拟工具：MiroFish 。
它不是传统意义上的“再问一次 ChatGPT”。MiroFish 更像一个把问题拆成可复盘流程的预测工作区：你先用自然语言描述要研究的场景，也可以补充文章、策略文档、市场信息、政策草案、研究笔记、小说片段等 seed material ；系统会提取实体和关系，构建 GraphRAG 知识图谱，再生成一批带有不同立场、记忆和行为逻辑的 AI agents ，让它们在模拟环境里互动，最后输出一份结构化的预测 / 场景报告。
我觉得它的价值主要在这些地方：
* 不是单模型直接给结论，而是通过多智能体互动来观察分歧、传播路径和群体反应。 * 输入门槛低，像聊天一样描述问题，复杂场景可以再加 PDF 、Markdown 、TXT 或纯文本资料。 * 输出不是一句“会/不会”，而是报告：可能路径、关键假设、风险点、时间线、反对意见和后续验证建议。 * 适合产品发布、涨价、品牌舆情、政策讨论、市场叙事、投资研究、创作分支等有人群反应和利益冲突的场景。 * 可以在结果出来后继续追问，了解某个模拟角色为什么这样反应，也可以修改一个假设后重跑，比较两次结果差异。 * 开源项目可自托管，适合希望自己控制模型、数据和运行环境的开发者；也有 hosted workspace ，适合不想自己配环境的人。
我会把 MiroFish 理解成“决策前的沙盘推演”，而不是保证未来一定发生的工具。它更适合回答这类问题：
“如果我把产品价格提高 20%，不同用户群会怎么反应？” “某个政策进入公共讨论后，支持和反对的声音可能怎样扩散？” “一个新产品发布后，市场叙事会偏向乐观、观望还是质疑？”
如果你正在做产品、市场、舆情、政策、研究或内容创作，想在行动前先把可能反应推演一遍，可以试试 MiroFish： https://mirofish.work/
也欢迎拍砖，尤其想听听大家觉得“多智能体模拟”在真实决策里最该解决哪一类问题。
Key Quotes
> MiroFish is more like a prediction workspace that breaks down problems into reviewable processes.
> It is better suited to answer questions like: If I increase the product price by 20%, how will different user groups react?
Tags
AI Agent
GraphRAG
Scenario Prediction
Decision Support
Open Source Tool
Related Articles
* Sharing My Multi-Project AI Terminal Codux: Unified Management of 8 CLIs like Codex/Claude Code by Project, Open Source * Understanding Hermes in One Article: How the New Top-Tier Agent Self-Evolves from Experience * Redefining Skill Development: A Step-by-Step Guide & One-Stop Development Assistant Launch * From OpenClaw to FastClaw: How to Design a Great Multi-Agent Architecture * Skill: Tips on How Vibe Coding Improves Code Quality and Efficiency * OpenClaw and Hermes: A Comprehensive Review of AI Agent Architecture from Source Code * Investing 10,000 RMB in AI: A Year of Tools, Practices, and Reflections * Heart Bleeding: My 32GB 16-inch MacBook Pro 2019 Suddenly Died that bricked after a system update, ultimately diagnosed as physical damage to the iNAND inside the T2 chip, along with subsequent solutions and data recovery notes.") * Harness Is Not the Goal, Knowledge Is the Moat — Knowledge Accumulation Practices of an AI Engineering Delivery Team * vibe-coding-template: After Losing a Codex Conversation File, I Organized a Set of Agent Long-Term Collaboration Templates to Quickly Embed AGENTS.md, Task Prompts, Code Review, Knowledge Explanations, and Web-Search Workflows into Projects
Make your daily reading actually fit you.A daily brief built from the sources you follow. Get started free HomeDiscoverWorld CupSettings