Install MariaDB MCP Server

Pick your client, copy the command, done.

claude mcp add mariadb-mcp -- python path/to/server.py

Environment Variables

Set these before running MariaDB MCP Server.

VariableDescriptionRequired

EMBEDDING_PROVIDERProvider for embeddings (openai, gemini, or huggingface)No

OPENAI_API_KEYAPI key for OpenAI embeddingsNo

GEMINI_API_KEYAPI key for Gemini embeddingsNo

HF_MODELModel name for HuggingFace embeddingsNo

MCP_READ_ONLYEnforces read-only mode for SQL queriesNo

Once configured, MariaDB MCP Server gives your AI agent access to:

list_databasesLists all accessible databases.

list_tablesLists all tables in a specified database.

database_name

get_table_schemaRetrieves schema for a table including columns, types, and keys.

database_nametable_name

get_table_schema_with_relationsRetrieves schema with foreign key relations for a table.

database_nametable_name

execute_sqlExecutes a read-only SQL query.

sql_querydatabase_nameparameters

create_databaseCreates a new database if it does not exist.

database_name

create_vector_storeCreates a new vector store table for embeddings.

database_namevector_store_namemodel_namedistance_function

delete_vector_storeDeletes a vector store table.

database_namevector_store_name

list_vector_storesLists all vector stores in a database.

database_name

insert_docs_vector_storeBatch inserts documents and metadata into a vector store.

database_namevector_store_namedocumentsmetadata

search_vector_storePerforms semantic search for similar documents using embeddings.

database_namevector_store_nameuser_queryk

After setup, try these prompts with your AI agent:

→List all the databases available on the server.

→Show me the schema for the 'users' table in the 'production' database.

→Execute a query to find all users who signed up in the last 30 days.

→Search the 'knowledge_base' vector store for documents related to 'database security best practices'.

→Create a new vector store named 'product_embeddings' in the 'catalog' database.

Prerequisites & system requirements

An MCP-compatible client (Claude Code, Cursor, Windsurf, Claude Desktop, or Codex)
Python 3.8+ with pip installed

Save the docs, env vars, and workflow around MariaDB MCP Server in Conare so Claude Code, Codex, and Cursor remember it next time.