SQLite KG Vec MCP

The SQLite KG Vec MCP server is a purpose‑built knowledge‑management platform that turns unstructured documents into a richly connected graph of entities and relationships, all while providing fast semantic search through vector embeddings. By exposing its functionality over the Model Context Protocol, it allows AI assistants such as Claude to treat document ingestion and graph querying as first‑class tools in a conversational workflow. This eliminates the need for developers to build custom pipelines or maintain separate search services, streamlining the creation of intelligent applications that can reason about and retrieve information from large corpora.

At its core, the server performs three complementary tasks. First, it parses incoming documents and automatically extracts nodes (concepts, people, places) and edges (relationships such as “author of” or “located in”) using a pluggable knowledge‑extractor. Second, it persists this graph in a lightweight SQLite database, preserving provenance and enabling transactional updates. Third, it builds vector representations of nodes and documents with any supported embedding provider (OpenAI, Ollama, HuggingFace, etc.) and indexes them with an HNSW search layer. The result is a searchable knowledge graph that can answer complex queries like “Show me all research papers authored by people from MIT” or “Find documents related to quantum computing that mention ‘entanglement’.”

Key capabilities include:

• Hexagonal Architecture: A clean separation of domain logic, ports, and adapters ensures that the core knowledge‑graph model remains framework‑agnostic and easily testable.
• Multiple Adapters: Developers can swap out SQLite for another persistence layer, switch embedding providers on the fly, or replace the LLM behind the extractor without touching business rules.
• Semantic Search: Vector‑based similarity search provides near‑real‑time retrieval of relevant documents or entities, even when keyword matching would fail.
• MCP Integration: The server implements the MCP interface, exposing resources such as , , and that AI assistants can invoke directly within a conversation, enabling dynamic data retrieval without leaving the chat context.

Real‑world use cases span academic research assistants that surface related papers, corporate knowledge bases that surface policy documents based on employee queries, and chatbot backends that can reference a company’s internal documentation in real time. By coupling document ingestion, graph storage, and vector search into a single, protocol‑ready service, the SQLite KG Vec MCP empowers developers to build AI‑enhanced knowledge workers with minimal friction and maximum flexibility.