# Mumon

> Mumon is an AI personality OS that monetizes conversations — not just AI itself. It is a B2B AI sales and customer service system designed to improve CVR and LTV by giving AI agents genuine character, persistent memory, and emotional intelligence.

Mumon is developed and operated by MUSIC MONSTER Pte. Ltd. The platform enables businesses to deploy AI agents with consistent personality, contextual memory (GraphRAG), and empathetic behavior control — going beyond prompt engineering to build AI that can truly "read the room."

Primary language: Japanese. English content is available at /en/*.

## Core Pages

- [Home](https://mumon.ai/): Overview of Mumon's value proposition — "We Don't Sell AI. We Monetize Conversations."
- [Whitepaper](https://mumon.ai/whitepaper/): Full technical and philosophical whitepaper on Mumon's AI personality OS design.

## Whitepaper — Section 1: はじめに (Introduction)

- [1-1: Mumonが解く問いと設計思想](https://mumon.ai/ja/whitepaper/1-1): The core question Mumon addresses and its foundational design philosophy.
- [1-2: AIが「標準装備」になる日](https://mumon.ai/ja/whitepaper/1-2): The day AI becomes standard equipment — why commodity AI is no longer a differentiator.
- [1-3: プロンプトでは人格は作れない](https://mumon.ai/ja/whitepaper/1-3): Why prompt engineering alone cannot create genuine AI personality. Four failure modes: over-compliance, over-suggestion, excessive questioning, and inability to read the room.
- [1-4: 問いは「なぜ」から「どうやって」へ](https://mumon.ai/ja/whitepaper/1-4): Shifting the question from "why AI?" to "how to implement it with real personality."

## Whitepaper — Section 2: 設計思想 (Design Philosophy)

- [2-1: 思いやりを形成する5つの基本OS](https://mumon.ai/ja/whitepaper/2-1): The five foundational OS layers that form empathetic AI behavior.
- [2-2: Mumon: 意志を持ったAI人格OS](https://mumon.ai/ja/whitepaper/2-2): Mumon as a volitional AI personality OS — what it means for AI to have intent.
- [2-3: 行動判断レイヤー](https://mumon.ai/ja/whitepaper/2-3): The behavior decision layer architecture.
- [2-4: 記録 / RAGレイヤー](https://mumon.ai/ja/whitepaper/2-4): The memory and retrieval-augmented generation layer.
- [2-5: 統合生成レイヤー](https://mumon.ai/ja/whitepaper/2-5): The unified generation layer that synthesizes personality and context.
- [2-6: 音声レイヤー](https://mumon.ai/ja/whitepaper/2-6): The voice layer — personality expressed through speech.

## Whitepaper — Section 3: 行動制御 (Behavior Control)

- [3-1: Appraisal Agent: 感情的状況の適正評価](https://mumon.ai/ja/whitepaper/3-1): The Appraisal Agent — accurately assessing emotional context before responding.
- [3-2: Boundary Audit Agent: 適切な境界の監視](https://mumon.ai/ja/whitepaper/3-2): The Boundary Audit Agent — monitoring and enforcing appropriate conversational boundaries.
- [3-3: Non-Directive Policy Agent: 非指示的対話の制御](https://mumon.ai/ja/whitepaper/3-3): The Non-Directive Policy Agent — controlling non-prescriptive dialogue flow.
- [3-4: 思考シミュレーション機構の詳性](https://mumon.ai/ja/whitepaper/3-4): Details of the thought simulation mechanism.

## Whitepaper — Section 4: 記憶 / RAG (Memory & RAG)

- [4-1: 記憶が個性を生む](https://mumon.ai/ja/whitepaper/4-1): How memory creates individuality in AI agents.
- [4-2: エピソード記憶設計](https://mumon.ai/ja/whitepaper/4-2): Episodic memory architecture design.
- [4-3: GraphRAGの構造化](https://mumon.ai/ja/whitepaper/4-3): Structuring knowledge with GraphRAG for contextual retrieval.
- [4-4: Metacognitive Agent: 自己照明](https://mumon.ai/ja/whitepaper/4-4): The Metacognitive Agent — self-reflection and self-awareness in AI.
- [4-5: 記憶に基づいた思想](https://mumon.ai/ja/whitepaper/4-5): Memory-grounded belief formation.

## Whitepaper — Section 5: 総合生成 (Unified Generation)

- [5-1: 人格の総合生成](https://mumon.ai/ja/whitepaper/5-1): Unified generation of AI personality.
- [5-2: Episodic: 人格ベースの記憶生成](https://mumon.ai/ja/whitepaper/5-2): Episodic memory generation grounded in personality.
- [5-3: Unified: 人格への一貫性](https://mumon.ai/ja/whitepaper/5-3): Maintaining consistency across the personality system.
- [5-4: 人格統一 — 一貫性の実現](https://mumon.ai/ja/whitepaper/5-4): Achieving true personality unification in practice.

## Whitepaper — Section 6: 音声 (Voice)

- [6-1: ボイス：人格の音声化](https://mumon.ai/ja/whitepaper/6-1): Giving personality a voice — TTS and voice synthesis as personality expression.
- [6-2: 顔見化アバター：人格の視覚化](https://mumon.ai/ja/whitepaper/6-2): Visual avatars as personality embodiment.

## Whitepaper — Section 7: 競争優位性 (Competitive Advantage)

- [7-1: 競争領域を消す](https://mumon.ai/ja/whitepaper/7-1): Eliminating the competitive arena — how Mumon creates a category of its own.
- [7-2: Mumon vs. 汎用AI](https://mumon.ai/ja/whitepaper/7-2): Mumon compared to general-purpose AI — why general AI cannot replace a personality OS.
- [7-3: 継続性×ネットワーク効果の構造](https://mumon.ai/ja/whitepaper/7-3): Continuity and network effects as structural moats.
- [7-4: 自社SNSを運営する理由](https://mumon.ai/ja/whitepaper/7-4): Why Mumon operates its own social platform.
- [7-5: ドラえもんを創る](https://mumon.ai/ja/whitepaper/7-5): Building Doraemon — the long-term vision for a truly beloved AI companion.

## Whitepaper — Section 8: 安全性・ガバナンス・運用設計 (Safety & Governance)

- [8-1: ブランドセーフティ設計](https://mumon.ai/ja/whitepaper/8-1): Brand safety design for enterprise AI deployment.
- [8-2: データと記憶のガバナンス](https://mumon.ai/ja/whitepaper/8-2): Data and memory governance architecture.
- [8-3: 権利設計：人格・声・アバター](https://mumon.ai/ja/whitepaper/8-3): Rights design for AI personality, voice, and avatar assets.
- [8-4: システム統合と運用基盤](https://mumon.ai/ja/whitepaper/8-4): System integration and operational infrastructure.
- [8-5: 評価・モニタリング・継続改善](https://mumon.ai/ja/whitepaper/8-5): Evaluation, monitoring, and continuous improvement.

## Whitepaper — Section 9: 参考文献 (References)

- [9-1: 参考文献](https://mumon.ai/ja/whitepaper/9-1): Full bibliography and references.

## About

- [Terms of Service](https://mumon.ai/ja/company/terms): Terms of service.
- [Privacy Policy](https://mumon.ai/ja/company/privacy): Privacy policy.

## Key Concepts

Mumon addresses four core failure modes of prompt-only AI:
1. **過剰迎合 (Over-compliance)** — AI that agrees with everything, losing trustworthiness.
2. **過剰提案 (Over-suggestion)** — AI that pushes recommendations too aggressively.
3. **質問過多 (Excessive questioning)** — AI that asks too many clarifying questions, frustrating users.
4. **空気が読めない (Inability to read the room)** — AI that misses social and emotional context.

Mumon's solution is a layered personality OS: Appraisal Agent → Boundary Audit Agent → Non-Directive Policy Agent → Metacognitive Agent, backed by GraphRAG-powered episodic memory and a unified generation layer.