What is the difference between listing tables and describing a table?

Listing returns names only. Describing returns structure: column names, types, and often indexes and keys. An agent can plan a query from a name list but writes a correct one far more reliably when it can read the columns, which is why servers like Neon and Airtable expose a describe tool alongside the list.

Do these work on NoSQL databases?

Yes, with a looser meaning of table. MongoDB lists databases and collections, and Neo4j returns a graph schema of labels and relationship types. The shape differs from SQL but the purpose is the same: let the agent learn the structure before it queries.

MCP servers that can list database tables

6 verified servers expose a tool that can list and inspect database tables

Before an agent can write a useful query it needs to know what is there: which tables exist, what the columns are called, how they relate. That discovery step is its own tool on most database MCP servers, separate from running the query itself.

These verified servers expose schema introspection, so an agent can list tables and read their structure before it touches a row of data.

Top pick

Supabase

Supabase (community)

Community

Connects agents to a Supabase project: run SQL, inspect schema, read logs, and manage edge functions.

databases2,710

Tool:

list_tables

Returns the tables in your Supabase Postgres in one call, which pairs naturally with its execute_sql tool: list first, then query what you found.

Pick 2

ClickHouse

Official

ClickHouse's official MCP server lets agents list databases and tables and run read-only SQL against a ClickHouse cluster.

databases793

Tools:

list_databases
list_tables

Walks the hierarchy from databases down to tables, which fits ClickHouse deployments where one server hosts many analytical databases.

Pick 3

Neon

Official

Neon's official MCP server lets agents create projects and branches, run SQL, and drive safe schema migrations on serverless Postgres.

databases606

Tools:

get_database_tables
describe_table_schema

Goes beyond a name list: describe_table_schema returns column types and structure, so an agent can build a correct query without a trial-and-error round trip.

Pick 4

Airtable

Adam Jones (domdomegg)

Community

Maintained Airtable MCP server: let an agent inspect base schemas, then read, search, and write records, tables, fields, and comments.

productivity443

Tools:

list_tables
describe_table

list_tables then describe_table surface field definitions, which matter more in Airtable than in SQL because field types like linked records and attachments change how you read a row.

Pick 5

MongoDB

Official

MongoDB's official MCP server: query and manage databases, plus Atlas cluster administration.

databases1,039

Tools:

list-databases
list-collections

list-databases then list-collections walk a MongoDB cluster down to the collection level, the closest thing to a table list in a store with no fixed schema.

Pick 6

Neo4j

Official

Neo4j's official MCP server lets agents introspect a graph schema and run read or write Cypher against any Neo4j deployment.

databases248

Tool:

get-schema

Returns the graph schema (node labels, relationship types, properties), the equivalent of a table list for a graph database where the structure is the model.

What to know

Schema tools matter more than they look. An agent that can list tables and describe columns writes far fewer broken queries, because it stops guessing at names like user_id versus userId. The richer servers go past a name list: Neon's describe_table_schema returns column types and structure, which is what lets a model join two tables correctly on the first try instead of after three failures.

Schemas are stable, but re-reading the full structure of a large database every session burns tokens for facts that did not change. Map it once and keep the map, rather than walk the same catalog cold on every run.

Questions

What is the difference between listing tables and describing a table?: Listing returns names only. Describing returns structure: column names, types, and often indexes and keys. An agent can plan a query from a name list but writes a correct one far more reliably when it can read the columns, which is why servers like Neon and Airtable expose a describe tool alongside the list.
Do these work on NoSQL databases?: Yes, with a looser meaning of table. MongoDB lists databases and collections, and Neo4j returns a graph schema of labels and relationship types. The shape differs from SQL but the purpose is the same: let the agent learn the structure before it queries.