# Connect to a database SnoutData talks to your databases directly from your machine. Credentials are encrypted by your operating system's keychain and never leave the app. ## Add a connection 1. Open the **connections sidebar** (the activity bar on the left). 2. Choose **New connection**. 3. Pick a **driver**. SnoutData supports relational databases (MySQL/Aurora, MariaDB, PostgreSQL, SQL Server, Oracle, IBM Db2, SQLite, DuckDB, Snowflake, Amazon Redshift, SAP HANA), the document database **MongoDB**, and the vector database **Pinecone**. See [supported databases](https://docs.snoutdata.com/connections/drivers). 4. Fill in the connection details. Most databases ask for a host, port, database, username, and password. (For SQLite or DuckDB, choose a local database file instead. For Snowflake, enter your account identifier, username, password, and a warehouse/role. For SAP HANA, leave **Database** blank unless you want to name a specific tenant database. For MongoDB you can paste a `mongodb://` connection string. For Pinecone, enter an API key.) 5. Click **Test** to verify, then **Save**.  Once saved, expand the connection to browse its contents: databases, tables, and columns, or collections (MongoDB) and indexes (Pinecone). ## Connecting over SSH If your database is only reachable through a bastion host, enable the **SSH tunnel** options on the connection form and provide the jump host plus a key or password. See [SSH tunnels](https://docs.snoutdata.com/connections/ssh-tunnel) for details. ## Flagging production Mark a connection as **production** to turn on the destructive-statement guardrail, which asks for confirmation before running statements like `DELETE` or `DROP` without a `WHERE` clause. See [Security](https://docs.snoutdata.com/connections/security). ## Next step [Run your first query](https://docs.snoutdata.com/getting-started/run-a-query).