Knowledge Vault Search

Search a personal knowledge vault using hybrid semantic and keyword matching.

This tool provides an MCP (Model Context Protocol) server that allows you to search your personal knowledge vault using hybrid semantic and keyword matching. The server connects to the Fondu Knowledge Vault API to retrieve relevant information from your personal knowledge base.

- Hybrid Search: Combines semantic vector search with keyword matching
- Reranking: Uses reranking models to prioritize the most relevant results
- Flexible Authentication: Multiple authentication methods with priority-based resolution
- Production Ready: Comprehensive error handling and logging
- MCP Compatible: Works with Claude Desktop and other MCP clients
- SSE Transport: Server-Sent Events for real-time communication

- Python 3.8+
- Access to Fondu Knowledge Vault API
- Valid authentication token

git clone <repository-url> cd mcp_tools

- Set up virtual environment and install dependencies:

python3 -m venv .venv source .venv/bin/activate pip install -r requirements.txt

The server supports multiple authentication methods with the following priority order:

1. Explicit Parameter (Highest Priority)

Pass the token directly when calling the tool.

export FONDU_AUTH_TOKEN="your-token-here" # or export FONDU_API_TOKEN="your-token-here"

Create a config file at one of these locations:

- ~/.fondu/config.yaml(recommended)
- ~/.config/fondu/config.yaml
- config.yaml(in project directory)

fondu: auth_token: "your-token-here" base_url: "https://api.youfondu.com" server: host: "127.0.0.1" port: 8080 debug: true

- ~/.fondu/token
- ~/.config/fondu/token
- .fondu_token

source .venv/bin/activate python mcp_fondu_search_user_context/server.py --host 127.0.0.1 --port 8080

The server will start onhttp://127.0.0.1:8080with the following endpoints:

- /- Homepage
- /health- Health check
- /sse- Server-Sent Events endpoint for MCP
- /messages/- Message handling endpoint

To use with Claude Desktop, add this to your~/Library/Application Support/Claude/claude_desktop_config.json:

{ "mcpServers": { "knowledge_vault": { "command": "python", "args": ["/absolute/path/to/mcp_tools/mcp_fondu_search_user_context/server.py"], "env": { "FONDU_AUTH_TOKEN": "your-auth-token-here" } } } }

{ "mcpServers": { "knowledge_vault": { "command": "/absolute/path/to/mcp_tools/run.sh", "env": { "FONDU_AUTH_TOKEN": "your-auth-token-here" } } } }

Search your knowledge vault using hybrid semantic and keyword matching.

- query(string, required): Natural language query for semantic search and reranking
- auth_token(string, optional): Authentication token (if not set via env/config)
- keywords(string, optional): Specific terms to prioritize in keyword matching
- top_k(integer, optional): Number of results to return (default: 10)

Returns:A formatted string containing the most relevant results from your knowledge vault, including source information and metadata when available.

Found 3 relevant results in your knowledge vault: 1. Quantum computing uses quantum mechanical phenomena like superposition and entanglement to perform calculations... Source: quantum_computing_notes.md Metadata: {'tags': ['physics', 'computing'], 'date': '2024-01-15'} 2. The fundamental principle behind quantum algorithms is the ability to exist in multiple states simultaneously... Source: research_papers/quantum_algorithms.pdf Metadata: {'author': 'Dr. Smith', 'year': 2023}

Create a test script to verify the tool works:

import asyncio import sys sys.path.append('mcp_fondu_search_user_context') from server import gather_relevant_user_knowledge async def test_tool(): result = await gather_relevant_user_knowledge( query="machine learning algorithms", auth_token="your-token-here", top_k=5 ) print(result) asyncio.run(test_tool())

Example configuration files are provided:

- config.yaml.example- Server configuration template
- claude_desktop_config.json.example- Claude Desktop setup template

Copy these files and customize with your settings:

cp config.yaml.example ~/.fondu/config.yaml # Edit with your auth token and preferences

The server provides comprehensive error handling:

- Missing Auth Token: Clear error message with setup instructions
- API Errors: Graceful handling of network issues and API failures
- Invalid Tokens: Proper 403 error handling
- Debug Logging: Detailed logs for troubleshooting

- Standard error output (visible when running the server)
- /tmp/error_log.txt(fallback error logging)

- Production API:https://api.youfondu.com/v1/knowledge/search_knowledge_vault
- Protocol: HTTPS with Bearer token authentication
- Timeout: 30 seconds for API requests
- Format: JSON request/response

- Verify your token is valid and not expired
- Check token is properly set via environment variable or config file
- Ensure no extra whitespace in token files

- Verify internet connectivity
- Check if API endpoint is accessible:curl -I https://api.youfondu.com
- Ensure no firewall blocking outbound HTTPS

- Restart Claude Desktop after configuration changes
- Check that absolute paths are used in configuration
- Verify Python virtual environment is properly activated

- Ensure all dependencies are installed:pip install -r requirements.txt
- Check port 8080 is not already in use
- Verify Python 3.8+ is being used

To enable debug logging, set the environment variable:

export PYTHONPATH=/path/to/mcp_tools export DEBUG=1 python mcp_fondu_search_user_context/server.py

You can test different authentication methods:

# Test with environment variable export FONDU_AUTH_TOKEN="your-token" python test_auth.py # Test with config file echo "fondu:\n auth_token: your-token" > ~/.fondu/config.yaml python test_auth.py # Test with token file echo "your-token" > ~/.fondu/token python test_auth.py

python3 -m venv .venv source .venv/bin/activate pip install -r requirements.txt

python test_mcp_server.py python test_tool_functionality.py

- mcp_fondu_search_user_context/server.py- Main server implementation
- requirements.txt- Python dependencies
- run.sh- Convenience script for starting server
- config.yaml.example- Configuration template
- claude_desktop_config.json.example- Claude Desktop setup template

- fastapi>=0.109.2- Web framework
- uvicorn>=0.27.1- ASGI server
- httpx>=0.26.0- HTTP client
- mcp>=1.3.0- Model Context Protocol
- PyYAML>=6.0- YAML configuration support

Seerequirements.txtfor complete dependency list.

This server is ready for deployment to AWS App Runner. SeeDEPLOYMENT.mdfor detailed deployment instructions.
- Push code to your Git repository
- Create App Runner service pointing to your repository
- SetFONDU_AUTH_TOKENenvironment variable
- Deploy!

- ✅ App Runner configuration (apprunner.yaml)
- ✅ Docker support (Dockerfile)
- ✅ Health check endpoint (/health)
- ✅ Environment variable configuration
- ✅ Production-ready logging
- ✅ Auto-scaling support

The server can be deployed to any platform that supports:

- Python 3.8+
- Environment variables
- HTTP/HTTPS traffic on port 8080

- AWS App Runner ✅
- Docker containers ✅
- Traditional VPS hosting ✅

Search global news using natural language. Webz.io News Search API returns the most relevant articles and content, with filters for source, country, language, date, sentiment, and category.

An MCP server providing semantic search capabilities for APLCart data.

MCP server for Christian scholarship and research — scripture, Greek/Hebrew word data, cross-references, patristic texts, and semantic search,

Access and search EPUB ebook collections using semantic vector search.

Provides full-text and semantic search over structured and unstructured data using Azure Cognitive Search.

local-first semantic search in Lojban dictionaries

A knowledge base server that processes local documents (PDF, DOCX, TXT, HTML) and answers questions based on their content using similarity search.

A managed Retrieval-Augmented Generation (RAG) server using MCP, integrated with knowledge bases and OpenSearch.

Search 419,000+ space regulatory filings from the FCC, ITU, UNOOSA, and FAA-AST — semantic search, entity dossiers, spectrum holdings, launch licenses, and alerts.

A semantic search server for codebases using Qdrant, featuring intelligent GitHub issue and project management.

RAG-focused MCP subregistry and server to discover and route to retrieval-capable MCP servers with semantic search, filters, and explainable ranking.