# Supported databases SnoutData connects to your databases through native drivers. Each driver handles connecting, introspecting the schema, and running queries; the editor, results grid, and AI features work the same way across all of them. Databases fall into three kinds: **relational**, **document**, and **vector**. ## Relational Classic SQL databases. You write SQL and it runs as SQL. | Database | Notes | | --- | --- | | **MySQL / Amazon Aurora (MySQL)** | Standard MySQL protocol. | | **MariaDB** | MySQL-compatible. | | **PostgreSQL** | Includes most Postgres-compatible engines. | | **Microsoft SQL Server** | TDS protocol. | | **Oracle** | Oracle Database. | | **IBM Db2** | Db2 for Linux, UNIX, and Windows (LUW). The first time you connect to a Db2 database, SnoutData downloads its client (a one-time, ~75 MB download) so it isn't carried in the installer. | | **SQLite** | Local, file-based databases. Pick the `.db` / `.sqlite` file instead of a host and port (no server or credentials needed). | | **DuckDB** | Local, file-based analytical (OLAP) database. Pick the `.duckdb` / `.db` file instead of a host and port (no server or credentials needed). The first time you connect to a DuckDB database, SnoutData downloads its engine (a one-time download) so it isn't carried in the installer. | | **Snowflake** | Cloud data warehouse. Connect with your account identifier, username, and password, plus a warehouse (and optional role) instead of a host and port. The first time you connect, SnoutData downloads its client (a one-time, ~20 MB download) so it isn't carried in the installer. Password sign-in for now (key-pair and SSO are planned); read-leaning, so turn on **Read-only** unless you need to write. | | **Amazon Redshift** | Cloud data warehouse (Serverless or provisioned). Connects like PostgreSQL: host, port 5439, database, username, and password, with optional SSH tunneling. Redshift has no indexes or triggers, and its primary and foreign keys are declared but not enforced by the engine, so SnoutData shows them for navigation and query help without implying they are validated. Read-leaning, so turn on **Read-only** unless you need to write; schema sync is not available for Redshift. | | **SAP HANA** | SAP's in-memory column store, on-premise or HANA Cloud. Host and port (typically `315`, or 443 on HANA Cloud), with optional SSH tunneling. Leave **Database** blank to use the tenant your port points at, or name a tenant database to select it. The first time you connect, SnoutData downloads its client (a one-time, ~5 MB download) so it isn't carried in the installer. | ## Document | Database | Notes | | --- | --- | | **MongoDB** | Collections show up as tables, with fields inferred by sampling documents. Query with SQL (transcribed to an aggregation pipeline) or write a native pipeline directly, edit documents in the grid, and manage collections and indexes. See [MongoDB](https://docs.snoutdata.com/databases/mongodb). | ## Vector | Database | Notes | | --- | --- | | **Pinecone** | Connect with an API key. Indexes show up as tables; query them with SQL (a metadata filter plus a result limit). See [vector databases](https://docs.snoutdata.com/databases/vector). | More drivers are on the roadmap. Each connection uses exactly one driver, chosen when you [create the connection](https://docs.snoutdata.com/getting-started/connect). ## One query language SQL is the common language across every kind of database. On a document or vector connection, SnoutData transcribes your SQL to the database's native form (a MongoDB aggregation pipeline or a Pinecone vector query), so the editor, results grid, and history work the same everywhere. For the full picture, see [querying beyond SQL](https://docs.snoutdata.com/databases/overview). ## Dialect awareness SnoutData adapts quoting, pagination, boolean handling, and similar details to each database's dialect, so the same features (the results grid, cell editing, generated SQL) behave correctly without you special-casing anything per vendor.