MCP Content Search is an MCP-based content indexing and search server built with LlamaIndex, ChromaDB, and a custom tool API.
Dynamic auto-fallback search (Local DB ➝ Web ➝ Auto-index)
Vector-based semantic search via LlamaIndex + ChromaDB
Real-time web search for Notion & Tistory
HTML crawling for sites without APIs
MCP tool exposure for seamless integration with AI clients
search_content — Dynamic search (local → web)
search_notion — Forced Notion-only search
search_tistory — Forced Tistory-only search
trigger_index_all_content — Run full indexing in background
get_index_status — Check indexing progress
mcp-content-search/
│
├── environments/
│ ├── config.py # AppConfig, NotionConfig, setup_chroma()
│ └── token.py # API keys, environment variables
│
├── core/
│ ├── models.py # IndexState, DocumentModel, statuses
│ └── utils.py # ContentHasher, helpers
│
├── indexing/
│ ├── converter.py # Convert Notion/Tistory → unified format
│ ├── manager.py # Handles index life-cycle
│ └── indexer.py # Index documents into Chroma
│
├── fetching/
│ ├── notion.py # Notion API client + processors
│ ├── tistory.py # Tistory RSS extractor + HTML parser
│ ├── fetcher.py # Unified fetcher for full indexing
│ └── web_searcher.py # Notion/Tistory real-time search
│
├── search/
│ ├── dynamic_search.py # Local-first auto-fallback search
│ └── service.py # Local Chroma search only
│
├── api/
│ └── tools.py # MCP tool handlers (search, indexing, status)
│
├── main.py # Application entry point
├── requirements.txt
└── README.md
🔧 environments/ — Configuration Layer
File
Description
Key Components
config.pyApplication settings
AppConfig, NotionConfig, setup_chroma()
token.pyEnv variable loader
NOTION_API_KEY, TISTORY_BLOG_NAME, etc.
🎯 core/ — Core Models & Utilities
File
Description
Key Components
models.pyData structures
DocumentModel, IndexStatusModel, IndexState
utils.pyUtility functions
ContentHasher
📚 indexing/ — Indexing Pipeline
File
Description
Key Components
converter.pyDocument transformation
DocumentConverter
manager.pyManager for indexing
IndexManager
indexer.pyIndex content.
ContentIndexer
🌐 fetching/ — Data Fetching Layer
File
Description
Key Components
notion.pyNotion integration
NotionAPIClient, NotionPageProcessor, NotionSearcher
tistory.pyTistory blog crawler
TistoryPostExtractor, TistorySearcher
fetcher.pyUnified fetch interface used for indexing
DocumentFetcher
web_searcher.pyUnified search interface for real-time web search
WebSearcher
🔍 search/ — Search Service
File
Description
Key Components
dynamic_search.pySemantic search via index DB or web, After web search, the results are indexed to index DB
DynamicSearchService
service.pySemantic search via index DB
SearchService
File
Description
Key Components
tools.pyMCP tool exposure
register_tools(), tool handlers
🚀 main.py — Application Entry Point
Function
Description
create_app()Initialize app components
mainStart MCP server
🔄 Architecture of MCP Tools (Client)
↓
[FastMCP]
↓ calls tool
[api/tools.py]
↓
DynamicSearchService → SearchService (local search)
↓ fallback
WebSearcher (Notion/Tistory)
↓
Background Indexing
↓
ContentIndexer → Chroma → LlamaIndex
Install dependencies:
pip install -r requirements.txt Start the MCP server:
The application will:
Load configuration
Initialize Chroma vector store
Prepare indexing and search services
Register MCP tools
Start the server
Ensure all required API keys (e.g., Notion, Tistory) are set in the environment.
ChromaDB directory is configured via AppConfig.
You can extend the system by adding new data fetchers or custom MCP tools.
When enough posts exist in the local index DB (found 3 results in local DB ) When local results are insufficient (Insufficient results (2/3), searching web... ) 
