StoryMC

A Kotlin-based Paper plugin that transforms Minecraft into a living, narrative-driven world through AI-integrated NPCs and voice synthesis.

Fully stable as a Human-in-the-loop (HITL) orchestration tool. It provides a robust narration engine and NPC framework managed by a human Game Master.

Note

This project is designed as the client-side implementation of a broader agentic simulation framework. The architecture is transitioning from a standalone monolith to a decoupled system, where this plugin handles environmental rendering and user interaction while a dedicated simulation engine governs autonomous AI behaviors and world-state logic.

What is Story?

Story is a production-ready Minecraft plugin that enables Dungeon Masters to create immersive narrative experiences with AI-powered NPCs. It allows facilitating D&D-style campaigns with persistent character memory, dynamic dialogue generation, and realistic voice synthesis.

• LLM-Powered Dialogues. NPCs generate contextually-aware responses using language models
• Voice Synthesis. Character-specific voices via ElevenLabs integration
• Persistent Memory. NPCs remember conversations, relationships, and story events
• Rich Character System. Complex personalities, backstories, relationships, and lore
• Location Awareness. NPCs understand their environment and react accordingly
• Multi-Player Support. Handle concurrent conversations across multiple players
• Character Schedules. NPCs follow daily routines and autonomous behaviors
• Plugin Ecosystem. Integrates with Citizens, MythicMobs, and other popular plugins

Roadmap: Evolving into a comprehensive simulation engine. The goal is to transition from manual GM scheduling to a fully autonomous architecture where AI agents manage their own schedules, behaviors, and world-state interactions with minimal human intervention.

Installation

Requirements

• Minecraft 1.21.1
• Paper 1.21.1
• Java 21
• OpenRouter API key (for LLM integration)
• ElevenLabs API key (for voice synthesis) (optional)

Dependencies

• Citizens - NPC management
• MythicMobs (optional) - Advanced mob behaviors
• PlaceholderAPI (optional)
• DecentHolograms (optional)
• ItemsAdder (optional)
• RealisticSeasons (optional)
• ReviveMe (optional)
• TheNewEconomy (optional)
• SuperVanish (optional)
• BetterHealthBar (optional)

Usage

Click here to get started.

Architecture

Current Production Stack

┌─────────────────────────────────────┐
│     Minecraft Server (Paper)        │
│  ┌───────────────────────────────┐  │
│  │      Story Plugin (Kotlin)    │  │
│  │  ┌─────────────────────────┐  │  │
│  │  │   Character System      │  │  │
│  │  │   - Personalities       │  │  │
│  │  │   - Relationships       │  │  │
│  │  │   - Memories            │  │  │
│  │  └─────────────────────────┘  │  │
│  │  ┌─────────────────────────┐  │  │
│  │  │   Lores & Locations     │  │  │
│  │  │   - Keywords & Context  │  │  │
│  │  └─────────────────────────┘  │  │
│  │  ┌─────────────────────────┐  │  │
│  │  │   LLM Integration       │  │  │
│  │  │   (OpenRouter API)      │  │  │
│  │  └─────────────────────────┘  │  │
│  │  ┌─────────────────────────┐  │  │
│  │  │   Voice Synthesis       │  │  │
│  │  │   (ElevenLabs)          │  │  │
│  │  └─────────────────────────┘  │  │
│  └───────────────────────────────┘  │
└─────────────────────────────────────┘

How It Works

1. Player Interaction: Player speaks in proximity to an NPC
2. Context Building: System gathers character personality, recent memories, location data
3. LLM Generation: OpenRouter API generates contextually-appropriate response
4. Voice Synthesis: ElevenLabs converts text to character-specific voice
5. Response Delivery: Audio and text delivered to players
6. Memory Update: Conversation stored for future context

Experimental Features

Vector Database Integration (In Development)

Working on semantic search capabilities using Python + vector databases:

• 13,000+ conversation embeddings from gameplay data
• Semantic memory retrieval (vs. simple recent + flagged)
• Improved context relevance for long-term character development

MCP Server

Model Context Protocol server for external LLM access to story data:

• Query character information programmatically
• Retrieve narrative context across sessions
• Enable advanced AI-assisted storytelling tools

LangGraph Multi-Agent Architecture

Exploring autonomous NPC behaviors with sophisticated cognitive architecture:

• Environmental perception systems
• Goal-driven decision making
• Multi-step reasoning for complex interactions
• Reduced manual DM orchestration

External Simulation in Go

Making minecraft solely a frontend layer, and the simulation layer is in a separate server. Using Redis Pub/Sub for communication.

Use Cases

Narrative Campaigns

Run D&D-style adventures with persistent NPC characters who remember player actions and develop relationships over time.

Living Worlds

Create settlements with NPCs that have daily schedules, relationships, and goals - making the world feel alive even when not directly interacting.

Interactive Storytelling

Let NPCs drive emergent narratives based on player choices and actions, rather than following predetermined scripts.

Educational/Roleplay Servers

Facilitate creative roleplay with AI-powered characters that stay in character and respond contextually.

Support

• Issues: GitHub Issues
• Discord: Discord Server

Note: Story is under active development. Production features are stable, but experimental features are subject to change.

License

This project is licensed under the Creative Commons Attribution-NonCommercial 4.0 International (CC BY-NC 4.0) License.