A Model Context Protocol server for MariaDB that enables AI assistants to query databases and perform semantic search using vector embeddings.
MCP MariaDB Server is a Model Context Protocol (MCP) implementation that provides a standardized interface for AI assistants to interact with MariaDB databases. It enables AI-driven data workflows by supporting both traditional read-only SQL operations and advanced vector-based semantic search capabilities for embedding-based document storage and retrieval.
Developers and data engineers building AI assistants (like those integrated with Claude Desktop, Cursor, or Windsurf) that need to query and analyze data from MariaDB databases, including those requiring semantic search over stored documents.
It offers a secure, extensible bridge specifically for MariaDB, emphasizing safe read-only operations by default while providing optional vector store management with support for multiple embedding providers (OpenAI, Gemini, HuggingFace), which is unique compared to generic database connectors.
MariaDB MCP (Model Context Protocol) server implementation
Open-Awesome is built by the community, for the community. Submit a project, suggest an awesome list, or help improve the catalog on GitHub.
Enforces read-only SQL operations (SELECT, SHOW, DESCRIBE) with parameterized inputs and configurable MCP_READ_ONLY mode, reducing injection risks and unintended writes.
Supports multiple embedding providers (OpenAI, Gemini, HuggingFace) for semantic search, allowing AI assistants to perform similarity searches on document vectors stored in MariaDB.
Provides tools to list databases, tables, and retrieve detailed schemas including foreign key relations, facilitating AI-driven data analysis and exploration.
Offers extensive environment-based settings for SSL/TLS connections, embedding providers, and authentication via FastMCP, enabling tailored security for production use.
Only allows read-only queries; write operations are not supported, which restricts use cases requiring data insertion, updates, or deletions without workarounds.
Lacks built-in authentication—users must manually configure external providers like GitHub OAuth or JWT using FastMCP, adding significant initial setup overhead.
Tied exclusively to MariaDB and its VECTOR type for embeddings, making migration to other databases impractical and limiting ecosystem flexibility.