SQLite
Connect to a local SQLite file path and browse or query it through the SQLite driver.
Database
Database Documentation
Use Unfour’s database client to inspect the state behind an API or server symptom with SQLite, PostgreSQL, and MySQL, or compatible MariaDB servers. openGauss uses PostgreSQL-protocol connections; table exports and SQL / Saved SQL keep the intended database context.
Supported databases
PostgreSQL
Connect with host, port, database, username, password credential, and the supported PostgreSQL catalog and schema flow.
openGauss
The openGauss preset uses PostgreSQL-protocol connections with server detection, table browsing, native DDL, and export improvements.
MySQL and MariaDB
Connect through the MySQL driver. MySQL workflows are verified; compatible MariaDB servers use the same path where their protocol and SQL behavior match.
Create a connection
SQLite
Choose SQLite, give the connection a name, enter the local database file path, choose read-only if needed, then Test and Save.
PostgreSQL
Choose PostgreSQL and enter host, port, database, username, and password. Test the target, review any SSL requirements, then Save.
MySQL
Choose MySQL and enter host, port, database, username, and password. Test the server, then Save. Use this path for a compatible MariaDB server.
Browse databases, schemas, and tables
- Select a saved connection and connect.
- Expand its database or catalog nodes. PostgreSQL can expose schemas inside the active catalog; MySQL exposes databases through the catalog list; SQLite uses its file database directly.
- Expand a table to review columns and primary-key markers.
- Open Structure to inspect available column details, indexes, foreign keys, and generated DDL information.
Browse and edit data
- Open a table to page through 50, 100, or 250 rows at a time.
- Use server-backed text filtering across available columns and sort by a selected column.
- Insert rows into writable tables. Update and delete require a table with primary-key columns so Unfour can identify the target row.
- Connections marked read-only do not expose row editing. Views and tables without a primary key do not allow update or delete.
- Changes are staged first. Preview the generated SQL, revert if needed, then confirm Apply; concurrency predicates help prevent overwriting a row that changed since it was loaded.
Run SQL
- Run All executes the full editor script, even when text is selected. Ctrl+Enter / Cmd+Enter runs All.
- Run Selected executes only highlighted SQL and is disabled without a selection. Ctrl+Shift+Enter / Cmd+Shift+Enter runs the selection.
- The full script is checked before execution. Statements run sequentially on one connection and stop at the first error.
- Each statement has its own result tab. Review returned columns and rows, affected rows, messages, and execution duration.
- Mutation and high-impact SQL requires an explicit confirmation before execution. Read-only connections reject writes.
- Stop prevents later statements from starting and waits for the current statement to finish; it does not roll back completed statements.
SQL and Saved SQL execution context
Unfour keeps the chosen database and schema when opening a query from the tree, running a script, reopening history, or opening Saved SQL. Discovering schemas no longer silently changes that selection. Saved SQL stores and restores its database and schema and appears under the appropriate database in the sidebar.
Before executing, review the connection, database, and schema shown in the query workspace. Saved context makes a query reusable; it does not make production SQL automatically safe.
Export table structure and data
- Open a table’s Export action.
- Choose Structure, Data, or Structure and Data. Structure uses SQL; data can use SQL, CSV, or JSON; combined export uses SQL.
- Choose a local destination file, run the export, and review the result. Only tables are supported; this is not a whole-database backup.
The MCP table-export tool also supports selected columns, filters, and row limits. Exporting data without a limit requires confirmation. Treat exported rows and DDL as potentially confidential.