mcp-vector-search

Category: Data Access Integration MCP Servers
Tags: mcp, semantic-search, vector-database, ai-integration, data-access

Description

mcp-vector-search is a Model Context Protocol (MCP) server that enables semantic data search using vector embeddings and similarity matching. It facilitates AI-powered, context-aware data retrieval, specifically designed for querying transcript segments stored in a Turso database. Users can ask questions and receive relevant transcript segments, with results ranked by semantic similarity.

Features

• Vector similarity search for transcript segments using embeddings
• Relevance scoring based on cosine similarity
• Transcript metadata in results (episode title, timestamps)
• Configurable search parameters:
  • Limit number of results (default 5, max 50)
  • Minimum similarity score threshold (default 0.5, range 0-1)
• Efficient database connection pooling for performance
• Comprehensive error handling
• Performance optimized for quick responses
• API endpoint: search_embeddings for semantic search
• Expects a Turso database with tables for embeddings and transcripts
• Complete development workflow (clone, install, build, run dev, publish)

API Details

• Endpoint: search_embeddings
  • Parameters:
    • question (string, required): Query text
    • limit (number, optional): Number of results
    • min_score (number, optional): Minimum similarity
  • Response: JSON array of results including episode title, segment text, start/end time, similarity score

Database Schema

Expects a Turso database with two tables:

• embeddings: id, transcript_id, embedding (vector)
• transcripts: id, episode_title, segment_text, start_time, end_time

Source Code

GitHub: spences10/mcp-embedding-search

Pricing

No pricing information provided; open source under MIT License.