The best Databases MCP servers
Let an agent read your data without you writing the query. These servers put a database behind the Model Context Protocol, so "which accounts churned last quarter?" is a question rather than a schema hunt and a hand-written join. Relational stores lead the list; the vector databases underneath them are the ones agents use for retrieval rather than reporting.
Official connectors first, then verified entries, then by how many tools they expose.
PostgreSQL
modelcontextprotocol
Direct PostgreSQL database access with read-only queries, schema inspection, and safe query execution.
MySQL
benborla29
MySQL database integration. Execute queries, manage schemas, and inspect database structure.
SQLite
anthropic
SQLite database operations including querying, analysis, and schema inspection. Perfect for local data management.
MongoDB
mongodb
MongoDB database integration for document operations, aggregation pipelines, and Atlas management.
Redis
modelcontextprotocol
Redis database integration for caching, pub/sub, and key-value operations through MCP.
Supabase
Supabase
Search the Supabase docs for up-to-date guidance and troubleshoot errors quickly. Manage organizations, projects, databases, and Edge Functions.
Neon Database
neon
Serverless PostgreSQL with branching. Create databases, run queries, and manage branches for development workflows.
PlanetScale
planetscale
Serverless MySQL platform. Branch databases like code, zero-downtime schema changes.
Turso
tursodatabase
Edge SQLite database. Distributed database with local replicas for low-latency reads.
BigQuery
LucasHild
Google BigQuery integration. Run SQL queries, manage datasets, and analyze large-scale data.
Snowflake
snowflake
Cloud data warehouse integration. Run queries, manage warehouses, and analyze large-scale data.
Elasticsearch
elastic
Full-text search and analytics engine. Index, search, and analyze large volumes of data in near real-time.
Prisma
prisma
Next-generation ORM for Node.js and TypeScript. Database access with type safety for PostgreSQL, MySQL, SQLite.
Airtable
domdomegg
Spreadsheet-database hybrid for structured data. Create, read, update records and manage bases programmatically.
MindsDB
mindsdb
AI query engine for federated data sources. Run ML predictions through SQL on any database.
Upstash
upstash
Serverless Redis and Kafka integration. Manage key-value data, message queues, and real-time data streams.
Chroma
chroma
Open-source vector database for AI applications. Store and retrieve embeddings for RAG and semantic search.
Pinecone
pinecone
Vector database for semantic search and RAG. Store, query, and manage vector embeddings at scale.
Weaviate
weaviate
Vector search engine with hybrid search capabilities. Combine vector and keyword search for optimal results.
Compare all 19 at a glanceShowHide
| Server | Tools | Status |
|---|---|---|
| PostgreSQL | 6 | Archived |
| MySQL | 7 | — |
| SQLite | 7 | Archived |
| MongoDB | 8 | Verified |
| Redis | 8 | Archived |
| Supabase | 8 | Verified |
| Neon Database | 7 | Verified |
| PlanetScale | 8 | — |
| Turso | 7 | — |
| BigQuery | 7 | — |
| Snowflake | 8 | Archived |
| Elasticsearch | 7 | Archived |
| Prisma | 7 | Verified |
| Airtable | 8 | Verified |
| MindsDB | 8 | — |
| Upstash | 8 | Verified |
| Chroma | 7 | Verified |
| Pinecone | 7 | Verified |
| Weaviate | 8 | Archived |
How we rank theseShowHide
Official connectors we host come first, then verifiedentries — ones where we hold the server’s own repository or the vendor’s own docs and have checked the link resolves — then the rest by how many agent-callable tools they expose. A listing on another directory doesn’t count as verification. We don’t publish install counts for third-party servers, because we have no way to measure them, and anything its maintainers have archived says so on the card. A dash in the Tools column means we haven’t listed that server’s tools yet, not that it has none.
Frequently asked questions
What is a database MCP server?
It exposes a database to AI models through the Model Context Protocol — usually as schema introspection plus a query tool — so a model can discover the tables itself and ask a real question of real rows, instead of you pasting a CSV into a chat window.
Can an agent write to or drop my tables?
Only if the credential you give it can. The control is the database role, not the server: create a read-only user, grant it the schemas you want reachable, and no prompt can exceed it. Our own connected-database feature works exactly this way, and we say plainly that a SQL classifier cannot make an owner credential safe.
Do I need one server per database?
With a config file, yes — each one is its own entry. That is the case for a hosted endpoint: BusinessMCP connects your database once and exposes it at a single URL alongside your analytics, CRM and revenue, so an agent reaches all of it through one connection rather than four.
Where do the vector databases fit?
Chroma, Pinecone, Weaviate and Upstash are retrieval infrastructure — an agent searches embeddings in them to find relevant context, then answers. A relational server answers questions about your business; a vector server helps the model find the passage that explains it. Most teams end up with both.
Connect PostgreSQL through one endpoint
Reach PostgreSQL and the rest of this list from Claude, ChatGPT, Gemini or Cursor without configuring each one. Free plan, no credit card.
Get started freeOr read how managed MCP hosting works first.