Databases and data warehouses: governed questions in plain language

Databases & warehouses connectors

• Amazon Athena

  Amazon Athena analytics in plain language: ask about Glue catalog tables on S3, Iceberg data, workgroups and the data each query scans.

  Status: Available
• Amazon Redshift

  Redshift MCP server for Claude and Copilot: ask questions of Amazon Redshift tables as your own Redshift user, with row-level security and visible SQL.

  Status: Available
• Cassandra

  A Cassandra dashboard in plain language: ask about keyspaces, tables, partitions and the application data in them, with the CQL shown for each answer.

  Status: Available
• ClickHouse

  ClickHouse MCP for governed questions: ask about events, logs and metrics in MergeTree tables, with row policies, the SQL and system.query_log in view.

  Status: Available
• Cosmos DB

  Cosmos DB analytics in plain language: ask about items in containers, partition keys, request unit consumption and 429 throttling, with the query shown.

  Status: Available
• Databricks

  Databricks MCP server for Claude and Copilot: ask questions of Unity Catalog tables as yourself, with row filters, column masks and the SQL shown.

  Status: Available
• Dremio

  Dremio AI questions over your lakehouse: ask views in the semantic layer, Iceberg tables and federated sources, with reflections and job history in view.

  Status: Available
• DynamoDB

  DynamoDB analytics in plain language: ask about items by partition and sort key, GSIs, consumed capacity and throttled requests, with each query shown.

  Status: Available
• Elasticsearch / OpenSearch

  Elasticsearch MCP questions over indices and data streams: logs, orders and search events answered with aggregations, under document-level security.

  Status: Available
• Google BigQuery

  BigQuery MCP server for Claude, ChatGPT and Copilot: ask questions of BigQuery datasets with your own Google account, IAM and row access policies.

  Status: Available
• IBM Db2

  Db2 reporting without hand-written SQL: ask about orders, claims and accounts in IBM Db2 tables, plus SYSCAT catalog and package cache facts.

  Status: Available
• InfluxDB

  InfluxDB dashboard questions in plain language: ask about sensor readings, server metrics and buckets, with the SQL, InfluxQL or Flux shown.

  Status: Available
• Microsoft Fabric

  Microsoft Fabric AI questions across lakehouses, warehouses, semantic models and eventhouses in OneLake, under each person's Entra ID permissions.

  Status: Available
• Microsoft SQL Server

  SQL Server MCP server for Claude and Copilot: ask questions of SQL Server or Azure SQL with your own login, row-level security and T-SQL shown.

  Status: Available
• MongoDB

  MongoDB MCP server for governed questions: orders, users and events in collections answered with aggregation pipelines, under each person's database role.

  Status: Available
• MySQL

  MySQL MCP server for Claude, ChatGPT and Copilot: plain-language questions on MySQL, RDS or Aurora MySQL as your own account, with the SQL shown.

  Status: Available
• Neo4j

  Neo4j MCP server questions over your graph: customers, accounts, suppliers and the paths between them, answered in Cypher under each person's role.

  Status: Available
• Oracle Database

  Oracle Database analytics without writing SQL: ask application schemas, views and materialized views, plus data dictionary and V$SQL facts.

  Status: Available
• PostgreSQL

  Postgres MCP server: ask business questions of PostgreSQL, RDS or Aurora as your own database role, with row-level security and read-only SQL.

  Status: Available
• SAP HANA

  SAP HANA analytics in plain language: ask calculation views, column tables and S/4HANA data in HANA, under analytic privileges, with the SQL shown.

  Status: Available
• SingleStore

  SingleStore analytics without writing SQL: ask live transactions and events in rowstore and columnstore tables, plus pipeline and workload facts.

  Status: Available
• Snowflake

  Snowflake MCP server for Claude, ChatGPT and Copilot: ask warehouse questions as yourself, with row access policies and the SQL shown on every answer.

  Status: Available
• Teradata

  Teradata analytics in plain language: ask the enterprise warehouse tables and views on Vantage, plus DBC dictionary and DBQL query log facts.

  Status: Available
• TimescaleDB

  TimescaleDB analytics in plain language: ask hypertables, continuous aggregates and joined PostgreSQL tables, with the SQL shown for each answer.

  Status: Available
• Trino / Starburst

  Trino analytics in plain language: ask Iceberg, Hive and relational catalogs through Trino or Starburst, under each user's access control rules.

  Status: Available
• Vertica

  Vertica AI questions over your columnar warehouse: business tables and views, projections and v_monitor query history, under Vertica access policies.

  Status: Available

Questions that cross these systems

• What was net revenue by region last quarter, using the finance mart's definition?
• Which customers' usage fell by more than half compared with their three-month average?
• Which tables in the analytics schema have not been loaded in the last 24 hours?

Questions and answers

Which databases does Nexatron answer today?

Snowflake, Google BigQuery, Databricks, Amazon Redshift, PostgreSQL, MySQL, Microsoft SQL Server and Azure SQL, and DuckDB through MotherDuck. Each page in this list says how its connection signs people in.

Is a database MCP server safe to give to Claude?

It depends on whose credentials it uses. A server that connects with one shared login lets every assistant user read whatever that login can. Nexatron signs each person in separately and runs only checked, read-only queries.