MCP + Claude Desktop
Connect a local MCP server to Claude Desktop app. Process PDFs automatically and use semantic search directly from Claude.
What You'll Learn
PDF Processing
Automatically extract text from PDFs and create embeddings
Claude Desktop Integration
Configure MCP servers for Claude Desktop app
Semantic Search
Query your documents using natural language
Source Tracking
Each result includes PDF source metadata
Workflow
Step 1: Environment Setup
# Navigate to the directory
cd Handson/MCP-Claude-Desktop
# Create virtual environment
python3 -m venv .venv
source .venv/bin/activate
# Install dependencies
pip install -r requirements.txt
# Verify installation
python3 -c "import faiss; import mcp; import PyPDF2; print('✅ Ready!')"Step 2: Add Your PDFs
Place your PDF documents in the data/PDF/ folder:
# Create the folder structure
mkdir -p data/PDF
# Copy your PDFs
cp /path/to/your/documents/*.pdf data/PDF/
# Your structure should look like:
# data/PDF/
# ├── document1.pdf
# ├── document2.pdf
# └── ...Step 3: Generate Vector Database
python3 generate_vector_db.py
# Expected output:
# ============================================================
# 🚀 FAISS Vector DB Generator from PDFs
# ============================================================
# 📂 Found 2 PDF file(s) in: data/PDF
# 📄 Processing: document1.pdf
# ✓ Extracted 15 chunks
# 📄 Processing: document2.pdf
# ✓ Extracted 12 chunks
#
# ✅ FAISS vector DB created successfully!
# 📁 Location: data/faiss_index
# 📦 Entries: 27 | Dimensions: 384Texts are split into ~1000 char chunks with 200 char overlap for better retrieval.
Edit the Claude Desktop configuration file:
~/Library/Application Support/Claude/claude_desktop_config.json
%APPDATA%\Claude\claude_desktop_config.json
~/.config/Claude/claude_desktop_config.json
{
"mcpServers": {
"vector-db": {
"command": "/ABSOLUTE/PATH/.venv/bin/python3",
"args": ["/ABSOLUTE/PATH/mcp_vector_server.py"],
"env": {
"PYTHONPATH": "/ABSOLUTE/PATH/MCP-Claude-Desktop"
}
}
}
}Important: Use absolute paths! Restart Claude Desktop (Cmd+Q) after editing.
Step 5: Use in Claude Desktop
In Claude Desktop, ask it to search your documents:
# Example prompt in Claude Desktop:
"Search 'quality principles' using search_vector_db with
index_dir='/path/to/data/faiss_index' and top_k=3"
# Expected response:
1) score=0.8234
[Source: company-handbook.pdf]
Critical TechWorks Quality (Quality, Reliability, and Resilience):
We ensure delivery of high-quality software...Learning Checklist
Quick Quiz
What must you do after editing claude_desktop_config.json?
"Server not found"
Check paths are absolute. Verify Python path
with which python3
"faiss.index missing"
Run python3 generate_vector_db.py
first
"BrokenPipeError"
Wait a few seconds after opening Claude Desktop before using tools
Check logs at:
~/Library/Logs/Claude/mcp-server-vector-db.log