Tabularis

What tabularis MCP server does

The tabularis MCP server exposes database access from the Tabularis desktop application to supported AI clients. Agents can inspect the schema available through Tabularis and run SQL queries without requiring a separate database-management interface. This makes it suited to tasks such as understanding table structure, checking data, and investigating query results from an AI conversation.

Tabularis is an open-source desktop SQL workspace rather than a standalone database service. It combines the MCP feature with a query editor, database explorer, SQL notebooks, a data grid, visual query-building tools, and visual EXPLAIN. The MCP server is therefore most relevant when the user already manages connections and queries in Tabularis.

How it works

Database connections are configured in Tabularis, and the built-in MCP server makes the relevant schema and query operations available to an AI client. The README identifies Claude, Cursor, and Devin as supported clients; it also notes that Devin was formerly called Windsurf. Queries run against databases connected through the desktop application, so the available data depends on those connections and their permissions.

The repository describes the MCP integration at a capability level: schema reading and query execution are the central agent functions. It does not document a separate hosted endpoint, authentication service, or standalone package for deploying the MCP server independently of Tabularis.

Tools and capabilities

The tabularis MCP server supports:

• Reading database schemas for agent context.
• Running SQL queries through configured connections.
• Working with the databases supported by Tabularis and its plugins.
• Using the same database workspace for both agent-assisted queries and manual SQL work.

Tabularis ships PostgreSQL, MySQL/MariaDB, and SQLite support, although the built-in PostgreSQL driver is deprecated in favor of the PostgreSQL plugin, which the application installs automatically. Other database systems are provided through plugins. The repository specifically lists DuckDB, ClickHouse, Redis, and Firestore among supported plugin-based options, but the exact MCP behavior for each plugin is not detailed in the provided material.

Limitations and notes

The tabularis MCP server depends on Tabularis and its configured database connections. It is not presented as a general-purpose database gateway with its own catalog of hosted connections. Database coverage is narrower out of the box than tools focused on dozens of drivers: three database drivers are built in, while additional coverage comes from official plugins.

The README does not specify individual MCP tool names, a transport configuration, environment variables, permission controls, or a separate command-line installation for the MCP feature. Install Tabularis using the platform package or installer documented by the project, then use its built-in MCP integration with a supported client. The project is licensed under Apache 2.0.