MCP MongoDB Server

MCP MongoDB Server

A high-performance Model Context Protocol (MCP) server that provides an efficient knowledge interface between Large Language Models (LLMs) and MongoDB. Optimized for small context windows (3k tokens), this server enables LLMs to retrieve and interact with domain-specific knowledge stored in MongoDB collections.

🚀 Features

• Dual-Server Architecture:

  • SSE-based MCP server for efficient LLM communication
  • RESTful CRUD API for database management

• Token Efficiency: Responses formatted for maximum information density within small context windows

• MongoDB Integration: Works with existing MongoDB collections with optimized query mechanisms

• Data Models for multiple domain-specific entities:

  • Novels
  • Chapters
  • Characters
  • Q&A Knowledge Base

• Python Scraper Integrations: Included as submodules for populating databases from various sources

🔧 Architecture

                    ┌───────────────────┐
                    │                   │
                    │     Language      │
                    │      Models       │
                    │                   │
                    └─────────┬─────────┘
                              │
                              │ MCP (JSON-RPC 2.0)
                              ▼
┌─────────────────────────────────────────────────────┐
│                                                     │
│  ┌─────────────────┐           ┌─────────────────┐  │
│  │                 │           │                 │  │
│  │   SSE Server    │◄────────►│   MCP Handler   │  │
│  │  (Port 3000)    │           │                 │  │
│  └─────────────────┘           └────────┬────────┘  │
│                                         │           │
│                                         ▼           │
│                                ┌────────────────┐   │
│  ┌─────────────────┐           │                │   │
│  │                 │           │  DB Services   │   │
│  │   CRUD API      │◄────────►│                │   │
│  │  (Port 3001)    │           └────────────────┘   │
│  └─────────────────┘                   │           │
│                                         │           │
└─────────────────────────────────────────┼───────────┘
                                          │
                                          ▼
                                 ┌─────────────────┐
                                 │                 │
                                 │    MongoDB      │
                                 │                 │
                                 └─────────────────┘

🏁 Getting Started

Prerequisites

• Rust (latest stable toolchain)
• MongoDB instance
• Python 3.7+ (for data scrapers)

Installation

1. Clone the repository with submodules:

   git clone --recursive https://github.com/SFBB/mcp-mongodb-novel-server.git
   cd mcp-mongodb-novel-server
   

2. Configure environment: Create a .env file in the project root:

   MONGODB_URI=mongodb://localhost:27017
   DATABASE_NAME=your_database
   PORT=3000  # Base port (MCP SSE server)
               # CRUD API will use PORT+1 (3001)
   

3. Build and run:

   cargo build --release
   cargo run --release
   

Upon successful startup, you'll see:

• MCP SSE endpoint available at http://127.0.0.1:3000/sse
• MCP POST endpoint available at http://127.0.0.1:3000/message
• CRUD API endpoints available at http://127.0.0.1:3001/api/...

Usage

You firstly need to run this server. Then configure your client based on JSON-RPC or SSE.

This is an exmaple setup for VSCode copilot client.

"mcp-mongodb-novel-server": {
    "type": "sse",
    "url": "http://localhost:3000/sse"
}  

🔌 Server Endpoints

1. MCP Server (SSE Protocol)

The MCP server provides two endpoints:

• SSE Endpoint: http://localhost:3000/sse
  • For establishing SSE connections to receive events
• POST Endpoint: http://localhost:3000/message
  • For sending JSON-RPC 2.0 formatted commands

Example MCP request (using POST to /message):

{
  "jsonrpc": "2.0",
  "id": "request-1",
  "method": "query_character",
  "params": {
    "character_id": "5f8e4c3b2a1d"
  }
}

2. CRUD API (REST)

The CRUD API provides RESTful endpoints for managing database content:

• Novels: /api/novels
• Chapters: /api/chapters
• Characters: /api/characters
• Q&A: /api/qa

Standard REST operations (GET, POST, PATCH, DELETE) are supported.

📊 Data Models

Novels

{
  "_id": "ObjectId",
  "title": "String",
  "author": "String",
  "summary": "String",
  "year": "Number",
  "tags": ["String"]
}

Chapters

{
  "_id": "ObjectId",
  "novel_id": "ObjectId",
  "title": "String",
  "chapter_number": "Number",
  "summary": "String",
  "content": "String",
  "word_count": "Number"
}

Characters

{
  "_id": "ObjectId",
  "novel_id": "ObjectId",
  "name": "String",
  "description": "String",
  "traits": ["String"],
  "relationships": [
    {
      "character_id": "ObjectId",
      "type": "String"
    }
  ]
}

Q&A

{
  "_id": "ObjectId",
  "question": "String",
  "answer": "String",
  "tags": ["String"],
  "novel_id": "ObjectId"
}

🤖 MCP Methods

The server supports the following MCP methods for LLM interaction:

Query Methods

Method	Description	Parameters
query_character	Get character details	{"character_id": "string"}
query_novel	Get novel metadata	{"novel_id": "string"}
query_chapter	Get chapter information	{"chapter_id": "string"} or {"chapter_number": number} or {"chapter_title": "string"}
query_qa_regex	Find Q&A entries by regex	{"regex_pattern": "string"}
query_chapter_regex	Find chapters by regex	{"regex_pattern": "string"}
query_character_regex	Find characters by regex	{"regex_pattern": "string"}

Update Methods

Method	Description	Parameters
update_chapter_summary	Update chapter summary	{"auth_token": "string", "chapter_id": "string", "summary": "string"}

📥 Data Population

This project includes Python scrapers as submodules to help populate your MongoDB:

Character Scraper

cd character-scraper
pip install -r requirements.txt
python src/scraper.py --config config/settings.json

Novel Scraper Library

cd scraper_library
pip install -r requirements.txt
python -m src.scrapers.scrape_syosetu --url <novel_url>

Available scrapers include:

• scrape_syosetu.py (Syosetu novels)
• scrape_69shu.py (69Shu novels)
• scrape_ximalaya.py (Ximalaya audio books)
• scrape_qa.py (Q&A content)
• scrape_x_timeline.py (Twitter/X timelines)
• And more...

🛠️ Development

Updating Submodules

# Update all submodules to their latest version
git submodule update --remote --merge

# Navigate to a specific submodule and pull changes
cd scraper_library
git pull origin main
cd ..
git add scraper_library
git commit -m "Update scraper_library"

Running Tests

cargo test

Code Style

This project follows Rust's standard code style guidelines:

# Check code formatting
cargo fmt -- --check

# Run linting
cargo clippy

Performance Tuning

For high-traffic deployments, consider:

1. MongoDB indexing for text search and relationships
2. Increase Tokio worker threads for CPU-bound operations
3. Configure MongoDB connection pooling

📄 License

This project is licensed under the MIT License - see the LICENSE file for details.