Features ›

Databases

Nine SQL engines and DynamoDB, with schema trees that go all the way down, a data grid that edits safely, and a query editor that tells you the truth about what ran.

Browse the data

Open a connection and the schema tree is on the left: tables, views, stored procedures, functions, indexes, foreign keys and triggers, each folder appearing only where that engine actually has the concept. Pick a table and its rows are in a real data grid, a page at a time.

Cells that are not text open in their own viewer: images as images, geometry on a map, anything else as hex
A badge says when a connection is not encrypted, rather than leaving you to remember
Edits in the grid are written with bound parameters, one transaction per apply, and each change must touch exactly one row

A PostgreSQL connection: the schema tree on the left and the customers table in a data grid on the right, 140 rows

The schema, down to the column

Expand a table to see its columns and their types, then its indexes and triggers. Foreign keys and primary keys are marked in the grid header too, so you can see how a table joins without opening its definition.

The orders table expanded in the schema tree: columns with their types, then Indexes and Triggers, beside its rows in the grid

Queries

The Query tab runs what you write on one dedicated session, so a transaction behaves like a transaction: BEGIN, an UPDATE and a ROLLBACK really roll back. Catalog reads and row counts use a separate connection, so browsing the tree never queues behind a long query or breaks an open transaction.

Cancel cancels on the server (on PostgreSQL and MySQL the query is stopped there, not just abandoned by the app). A result that hit the row limit says it was truncated, and a stream that fails part-way is reported as a failure, never as a shorter success.

Export any result as CSV, tab-separated, JSON, Markdown or SQL INSERT statements, streamed in batches
History of what you ran, per connection
Syntax highlighting and completion of table and column names

The Query tab: a revenue-by-city query with a join and GROUP BY, and its ten rows of results below

Every engine, built in

No driver downloads and no plugin marketplace. Each engine is compiled into the app.

Engine	Notes
PostgreSQL	Catalog-driven schema tree; server-side cancel.
MySQL & MariaDB	Catalog-driven schema tree; server-side cancel.
Microsoft SQL Server	2005 and later. SQL login or Microsoft Entra ID: browser sign-in with MFA, device code, Azure CLI or a service principal.
Oracle	8i and later.
SQLite	Indexes, foreign keys and triggers.
DuckDB	Indexes and foreign keys.
ClickHouse	Data-skipping indices, because the engine has no others.
Amazon Athena	External tables, so no index folders.
Amazon DynamoDB	Tables, keys and indexes; items read a page at a time because a scan is billed; PartiQL.

Safe to point at production

A connection can be marked read-only. Hydra then refuses any statement that could change data, in your own Query tab, in the grid, and from automation alike, using one classifier so the two can never disagree. A second setting, agents read-only, lets you write while scripts and AI agents can only read.

A database behind a bastion host connects through an SSH tunnel, everywhere: the tree, the grid, queries, and its own Test button.