Features

The MariaDB Enterprise MCP Server offers a comprehensive suite of tools, categorized into standard database operations, advanced vector functionalities, and workflow orchestration.

Standard Database Operations

These tools provide fundamental control and insight into your MariaDB environment. By default, operations are read-only (MCP_READ_ONLY = true) but can be configured for write access (MCP_READ_ONLY = false).

• list_databases: Discovers all accessible databases.

• list_tables: Enumerates all tables within a specified database.

• get_table_schema: Retrieves the detailed schema for a specific table, including column names, data types, keys, and default values.

• execute_sql: Executes read-only SQL queries like SELECT, SHOW, and DESCRIBE. Supports parameterized queries for enhanced security.

• create_database: Creates a new database if it does not already exist.

Harnessing the Power of Vectors: Advanced AI Functionality

The server’s integrated vector functionality enables semantic search and other embedding-based operations directly within your database.

Vector Store Management

• create_vector_store: Creates a new table optimized as a vector store. The schema includes columns for id, document, embedding (VECTOR type), and metadata (JSON). Users can specify the embedding model and distance function (e.g., cosine, euclidean) at creation.

• list_vector_stores: Lists all tables in a database that are identified as vector stores.

• delete_vector_store: Securely removes a vector store table.

Embedding and Search Operations

• insert_docs_vector_store: Inserts documents and associated metadata into a vector store. The server manages the generation of embeddings using a configured service.

• search_vector_store: Performs semantic similarity searches by generating an embedding for a user query and finding the 'k' most similar documents in the specified vector store.

Workflow Orchestration

The server exposes powerful orchestration endpoints that allow an AI agent to execute an entire RAG pipeline through a single, secure interface.

• Ingestion (/orchestrate/ingestion): Triggers the ingestion of documents into a specified vector store, including the chunking and embedding processes.

• Hybrid Search (/hybrid_search): Executes a search that combines semantic (vector) search with traditional keyword search to retrieve the most relevant document chunks.

• Generation (/orchestrate/generation): Executes a query against a set of documents, performing retrieval and generating a final, context-aware response from an LLM.

---

Tool Summary

Tool Name	Description	Category
list_databases	Discovers all accessible databases.	Standard Database Operations
list_tables	Enumerates all tables within a specified database.	Standard Database Operations
get_table_schema	Retrieves the detailed schema for a specific table.	Standard Database Operations
execute_sql	Executes read-only SQL queries.	Standard Database Operations
create_database	Creates a new database if it does not already exist.	Standard Database Operations
create_vector_store	Creates a new table optimized as a vector store.	Vector & AI Functionality
list_vector_stores	Lists all tables identified as vector stores.	Vector & AI Functionality
delete_vector_store	Securely removes a vector store table.	Vector & AI Functionality
insert_docs_vector_store	Inserts documents and metadata into a vector store.	Vector & AI Functionality
search_vector_store	Performs a semantic similarity search on a vector store.	Vector & AI Functionality
rag_ingestion	Triggers the full document ingestion pipeline.	Workflow Orchestration
rag_hybrid_search	Executes a combined semantic and keyword search.	Workflow Orchestration
rag_generation	Synthesizes retrieved information with the user's query to generate a final, context-aware response.	Workflow Orchestration