Tabularis: three built-in drivers, 26 shipped plugins and an MCP server in a Tauri shell

Three one-line installs and six downloadable artifacts

Three package managers cover the three desktop platforms, and the v0.30.0 release carries six files you can fetch by hand.

winget install Debba.Tabularis                                   # Windows
brew install --cask tabularis  # macOS
sudo snap install tabularis                                      # Linux

The release assets are tabularis_0.30.0_x64-setup.exe for Windows, tabularis_0.30.0_aarch64.dmg and tabularis_0.30.0_x64.dmg for macOS, tabularis_0.30.0_amd64.AppImage and tabularis_0.30.0_amd64.deb for Linux, and tabularis-0.30.0-1.x86_64.rpm. macOS gets one disk image per architecture, so an Apple silicon machine and an Intel machine do not share a file. Flatpak, Arch User Repository and winstall channels exist as well, alongside the direct installers. The interface is available in thirteen languages: English, Italian, Spanish, Simplified Chinese, Traditional Chinese, French, German, Japanese, Russian, Korean, Tagalog, Brazilian Portuguese and Vietnamese. The three built-in engines are PostgreSQL, MySQL/MariaDB and SQLite.

The built-in PostgreSQL driver is deprecated in favour of a plugin

PostgreSQL, MySQL/MariaDB and SQLite ship in the binary, and everything else is a plugin. The wrinkle is PostgreSQL itself: the built-in driver is deprecated in favour of the PostgreSQL plugin, which Tabularis installs automatically. That matters when you pin an app version, sit on a locked down network, or hand an install to a colleague. A connection you believed was bundled now depends on the separate PostgreSQL plugin, which lives in its own tabularis-postgresql-plugin repository.

Plugins speak JSON-RPC over stdio, so the language stops mattering

A plugin can be written in any language because the contract is JSON-RPC over stdio, and the project's own comparison marks that row as the difference from the Java and Eclipse plugins of DBeaver and the JavaScript plugins of TablePlus.

Coverage is published as shipped, claimed, scoped, coming soon or open

The README's database support section mirrors the driver and plugin coverage page on tabularis.dev and opens with a Shipped table. Shipped plugins named in the project description include DuckDB, ClickHouse, Redis and Firestore. Check that page for the current status of any other database before planning a migration.

The MCP server and Ollama put an agent and a local model in the same window

The built-in MCP server lets Claude, Cursor and Devin (formerly Windsurf) read your schema and run queries inside the app you already have open. The surface is narrow on purpose: schema reading plus query execution. What an agent can do is therefore bounded by those two operations and by whatever the database account behind the connection is allowed to do, which is a smaller blast radius than an agent with a shell. For text to SQL, the app lists local models through Ollama, and the AI features section is marked optional. The comparison table puts a cloud based assistant in the DBeaver column, so the two designs differ in where your schema text goes: the Tabularis row lists local models through Ollama.

Notebooks are a file format, which is why they survive a reinstall

SQL notebooks mix SQL cells with Markdown cells, let variables cross between cells, and can render charts, and the project's own comparison marks that row as missing in DBeaver CE, TablePlus and Beekeeper Studio. The unit of storage is a file with a .tabularis-notebook extension, and demo/notebook-showcase.tabularis-notebook sits in the repository, so a notebook can live in version control next to the queries it explains instead of inside an application profile. The demo directory also carries connections.json, docker-compose.yml, an init directory and generate-perf-sql.py, so a sample stack and a bulk SQL generator arrive with the source rather than with the installer.

Versions and changelogs are generated, so commit messages decide what you can read

The root package.json and the README download links both point at 0.30.0, and the tree carries release-please-config.json and .release-please-manifest.json next to CHANGELOG.md. The stable line is v0.30.0, and nightly-20261102-0116178 was published on 2026-11-02 as Nightly (v0.30.1-5), so a nightly install sits ahead of the stable release. The last push to the branch is dated 2026-11-02.

Breadth is the limit the project states about itself

Driver breadth is where this app stops, and the project says so plainly. The comparison, dated June 2026, counts 3 built-in drivers plus 26 driver plugins for Tabularis against 100+ for DBeaver CE, 20+ for TablePlus and roughly 10 for Beekeeper Studio, and the note beneath it says that if you need dozens of drivers you should use DBeaver, because Tabularis focuses on doing a few databases well. That note also warns that features in the other tools may have changed since June 2026, so treat the row as a snapshot rather than a verdict. The practical test is short. A team running PostgreSQL and Redis is inside the range today. A team that needs dozens of drivers is pointed to DBeaver by the README itself.

Editorial conclusion

Use Tabularis when your daily work is PostgreSQL, MySQL/MariaDB or SQLite and you want notebooks, local Ollama text to SQL and an agent connection in one window. Skip it if you need dozens of drivers, where the README itself points to DBeaver. Before installing, read the status of each database you actually run on the driver and plugin coverage page at tabularis.dev.

Frequently asked questions

Official sources

1. License: Apache-2.0
2. Project website
3. README
4. Releases
5. TabularisDB/tabularis on GitHub