TupleBase

Your database workspace, inside VS Code.

Explore schemas, write queries, inspect results, and revisit query history without leaving your editor.

TupleBase keeps connection configuration in a committable .tuplebase.json while credentials remain safely stored on your machine. Its built-in MCP server gives AI agents access to the same connections, with writes blocked by default.

TupleBase 0.1.0 is a pre-release. PostgreSQL is the engine enabled in this first public version.

Features

• Browse schemas from a connection explorer organized into project-defined groups.
• Run SQL where you write it using CodeLens actions, cmd+enter / ctrl+enter, or whole-file execution.
• Inspect results in a VS Code-themed grid with pagination and a structured detail view for each row.
• Keep configuration with the project in a comment-friendly .tuplebase.json that rejects embedded secrets.
• Store credentials locally through VS Code Secret Storage, with optional prompt-on-connect credentials and SSH tunnels.
• Control writes at the group or connection level before a statement reaches the database.
• Return to previous work through per-workspace query history and one-click reruns.
• Connect AI agents through the bundled, standalone MCP server.

Get started

1. Install the TupleBase pre-release from the VS Code Marketplace or Open VSX.
2. Open a folder in VS Code, select the TupleBase activity-bar icon, and choose Create Config.
3. Add a PostgreSQL connection from the explorer. TupleBase writes its non-secret settings to .tuplebase.json and prompts for the password when you first connect.
4. Open a .sql file and use the Run CodeLens or cmd+enter / ctrl+enter. TupleBase remembers the connection selected for that file.

A minimal configuration looks like this:

{
  "version": 1,
  "groups": {
    "development": {
      "readonly": true,
      "app-db": {
        "adapter": "postgres",
        "host": "localhost",
        "port": 5432,
        "database": "app",
        "user": "developer"
      }
    }
  }
}

Passwords and private keys do not belong in this file. TupleBase stores credentials through VS Code Secret Storage and also supports ${env:VARIABLE} references for non-secret configuration values.

Supported engines

Available in 0.1.0: PostgreSQL.

Adapters for MySQL and MariaDB, SQLite, SQL Server, ClickHouse, Cassandra, Neo4j, MongoDB, Elasticsearch and OpenSearch, Kafka, Redis, and DynamoDB are in development. See the database support matrix for their current status.

MCP for AI agents

TupleBase bundles a standalone Model Context Protocol server that uses the same connections and adapters as the extension. It gives compatible agents three tools:

• list_connections lists configured connections and their safety settings.
• inspect_schema explores database objects without issuing ad hoc discovery queries.
• run_query executes a statement and returns structured rows, columns, timing, and warnings.

Agent access is read-only by default. Allowing writes requires both an MCP server opt-in and a writable target connection. Run TupleBase: Show MCP Server Config to generate a client configuration, and see the MCP guide for setup and security details.

Security, privacy, and support

TupleBase does not collect telemetry or send usage data to TupleBase. It connects only to the database and SSH endpoints you configure.

• Report bugs and request features through GitHub Issues.
• Report vulnerabilities privately according to the security policy.
• For other questions, contact hello@tuplebase.app.

License

MIT. Third-party components and their licenses are listed in THIRD-PARTY-NOTICES.