MCP Servers Link LLMs to Nearly 30 Databases as Vendors Expand AI Access
Updated
Updated · InfoWorld · Jun 8
MCP Servers Link LLMs to Nearly 30 Databases as Vendors Expand AI Access
3 articles · Updated · InfoWorld · Jun 8
Summary
Vendor-backed MCP servers now let LLM tools query, update and administer databases through natural-language prompts, extending agentic access across SQL, NoSQL, graph, vector and warehouse systems.
Google’s open-source MCP Toolbox is the broadest option in the report, shipping prebuilt support for nearly 30 databases, while AWS, MongoDB, Neo4j, Redis, Snowflake, Supabase and Pinecone each offer official servers for their own platforms.
Capabilities vary by product: BigQuery’s remote server has limits on result size and processing time, Redis MCP is still local-only, Elastic routes access through its Agent Builder rather than raw APIs, and some tools remain pre-1.0 or deprecated.
The appeal is faster AI-assisted development—schema-aware querying, code generation, automation and debugging—but the report warns that prompt injection remains unresolved and recommends least-privilege access, manual approval for tool calls and internal MCP registries.