Using Sheeta AI
DB Connections
A project can hold any number of database connections. You pick the active connection in the Database sidebar, and tables, queries and imports use that connection.
Supported databases
The database type list in the connection form offers:
- SQLite and Turso SQLite: local database files. See SQLite and Turso.
- Postgres: any PostgreSQL server, including Amazon RDS. See Postgres.
- MySQL. See MySQL.
- MSSQL: Microsoft SQL Server. See SQL Server.
- BigQuery. See BigQuery.
- CloudSQL: Google Cloud SQL for PostgreSQL. See Google Cloud SQL.
- Supabase and Neon: hosted PostgreSQL with a built-in guide. See Supabase and Neon.
Add a connection
- In the Database sidebar, click Manage database connections. In an empty project you can also click Create Local DB for an instant SQLite database, or Add Connection.
- In Connection Management, click Add and choose the database type.
- Fill in the form. PostgreSQL-based types and MySQL also accept a connection URL on the URL tab: paste it and click Parse.
- Click Test Connection, then Save. Sheeta AI tests the connection again before saving and does not save a connection that fails.
Connection names must be unique within a project, ignoring upper and lower case. A new connection becomes the active connection.

Manage connections
Connection Management lists every connection in the project with a status dot: yellow while connecting, green when connected and red with the error when the connection fails.
- Click a connection to edit its settings.
- Click its test icon to test it, or the refresh icon to test all.
- Click its trash icon and confirm with Delete to remove it.
- Use Search connections... to filter the list.
From the Database sidebar you can also create, rename and delete schemas (datasets in BigQuery) and tables, add and rename columns, and duplicate, truncate or delete tables. Right-click a schema or table to see the options for your database.
Docker container tunnel
For PostgreSQL, MySQL and SQL Server running in a local Docker container whose port is not published, turn on Docker container tunnel. Enter the Docker container name in place of the host. Sheeta AI starts a temporary proxy container that listens only on localhost and forwards to the database container.

- Docker must be running on the same computer.
- The database container must be running.
- The first use downloads the small proxy image.
Sheeta AI does not support SSH tunnels.
Encryption
There is no SSL setting in the connection form. Sheeta AI turns on SSL automatically for known cloud hosts, such as Supabase, Neon, Amazon RDS, Azure, Render, Railway and Cloud SQL.
Where connections are stored
Connections belong to the project and are stored in its folder:
.sheeta/connections.jsonholds the host, port, database and other settings..sheeta/.envholds usernames, passwords and service account keys. This file is not encrypted.
Sheeta AI adds .env to .sheeta/.gitignore, so credentials stay out of Git. When a teammate clones the project, they see the same connections and enter their own credentials. Credentials never leave your computer, and the AI agent never receives them. Anyone with access to the project folder on your computer can read the .env file, so protect the folder as you would any file that contains passwords.