neo4j-semantic-search-mcp

# Semantic Search MCP Server **Universal, schema-agnostic** semantic search for **any** Neo4j knowledge graph. One tool. Works everywhere. No configuration needed. ## What It Does Searches across ALL vector indexes in your Neo4j database using semantic similarity. Returns anything that matches your natural language query above a threshold. Works with: - Any node type - Any relationship type - Any schema - Any knowledge graph with embeddings ## Installation ### 1. Install Dependencies ```bash pip install -r mcp_servers/semantic_search/requirements.txt ``` ### 2. Configure Environment Variables ```bash # Required export OPENAI_API_KEY="sk-..." export NEO4J_PASSWORD="your-password" # Optional (defaults shown) export NEO4J_URI="bolt://localhost:7687" export NEO4J_USER="neo4j" export EMBEDDING_MODEL="text-embedding-3-small" ``` ### 3. Register with Claude Desktop Add to `~/Library/Application Support/Claude/claude_desktop_config.json` (macOS): ```json { "mcpServers": { "semantic-search": { "command": "python", "args": ["-m", "mcp_servers.semantic_search.server"], "cwd": "/path/to/your/e2r-rd", "env": { "OPENAI_API_KEY": "sk-...", "NEO4J_PASSWORD": "your-password" } } } } ``` ### 4. Restart Claude Desktop ## Usage ### The Tool: `semantic_search` One simple tool with three parameters: ```python semantic_search( query: str, # Natural language query limit: int = 10, # Max results (max: 100) threshold: float = 0.7 # Min similarity (0-1) ) ``` **Examples:** ``` semantic_search("database failures") semantic_search("kubernetes scaling", limit=20) semantic_search("cache problems", threshold=0.8) ``` **Returns:** ``` Found 5 results matching 'database failures': 1. **Redis Cache** [Service] (similarity: 0.923) - mentioned 15 times 2. **PostgreSQL** [Service] (similarity: 0.887) - mentioned 12 times 3. **cache timeout** [Action] (similarity: 0.845) - mentioned 8 times 4. **d…