Embedding MCP Server

What is Embedding MCP Server?

Embedding MCP Server is a Model Context Protocol (MCP) server implementation powered by txtai that provides semantic search, knowledge graph capabilities, and AI-driven text processing through a standardized interface. It includes a knowledge base builder tool for creating and managing knowledge bases from documents.

How to use Embedding MCP Server?

Install with uv pip install kb-mcp-server or pip install kb-mcp-server, then build a knowledge base using kb-build --input /path/to/documents --config config.yml and start the server with kb-mcp-server --embeddings /path/to/knowledge_base. Alternatively, use uvx --from [email protected] kb-mcp-server --embeddings /path/to/knowledge_base for no-install execution. The server can be configured via command-line arguments (e.g., --host, --port, --transport) or environment variables.

Key features of Embedding MCP Server

• Unified vector database combining vector indexes, graph networks, and relational databases.
• Semantic search based on meaning, not keywords.
• Knowledge graph integration for automatic building and querying.
• Portable knowledge bases as compressed archives (.tar.gz).
• Local-first architecture; no data sent to external services.
• Extensible pipeline system for text, documents, audio, images, and video.

Use cases of Embedding MCP Server

• Semantic search across document collections using meaning-based queries.
• Building and querying knowledge graphs from unstructured text data.
• Running AI-driven text processing pipelines (summarization, extraction).
• Sharing portable knowledge bases as .tar.gz archives between systems.

FAQ from Embedding MCP Server

What is txtai and why is it used?

txtai is an all-in-one embeddings database for RAG that provides semantic search, knowledge graph construction, and language model workflows. This server leverages txtai as its core embedding and search engine.

What are the runtime requirements?

Python 3.10 or newer is recommended. Installation can be done via uv, pip, or conda. The server runs locally and does not require connection to external services.

Where does the knowledge base data live?

Knowledge bases are stored locally on the file system—either as plain folders or as portable .tar.gz archives. The server loads them from a path specified via the --embeddings argument.

What transports are supported?

The server supports both sse and stdio transports, configurable with the --transport argument. The default transport is stdio.

Is authentication or authorization supported?

The README does not mention any authentication or authorization mechanisms for the server.