Databases
functionalThe 30 engines SQLly connects to today, from SQL Server and PostgreSQL to DuckDB, ClickHouse, Oracle and Redis, which tools work on each, and how to connect.
SQLly understands 38 database engines. Thirty connect today through native Rust drivers, grouped around the seven wire protocols and file formats those engines actually speak, and one of them is not a SQL database at all. The remaining eight are fully profiled: SQLly knows their capabilities, quoting, row limits, and type storage, with live connectors still to come.
Each first-party engine also has its own page, what works on it, how to connect, and what does not apply, starting from Databases SQLly connects to.
Connect today, native drivers
Wire-compatible engines share a driver: PostgreSQL-protocol engines use the PostgreSQL connector, the MySQL family shares MySQL's, and the Redis-compatible stores share the Redis connector. Compatibility stays explicit, so a hosted or distributed engine says which first-party wire it uses and what differs.
TDS - SQL Server family
| Engine | SQL dialect | Notes |
|---|---|---|
| Microsoft SQL Server | T-SQL | First-party. Default schema dbo, [bracket] quoting, SELECT TOP n. |
| Azure SQL Database | T-SQL | Managed SQL Server; Entra ID sign-in and Azure discovery supported. No SQL Agent. |
PostgreSQL wire
| Engine | Compatible with | Notes |
|---|---|---|
| PostgreSQL | - | First-party. Default schema public, "double-quote" quoting, LIMIT, ILIKE, RETURNING, ON CONFLICT upsert. |
| CockroachDB | PostgreSQL | Distributed SQL over the pg wire. No VACUUM, so the Maintenance menu is unavailable; no Procedure Mode or Security editor. |
| YugabyteDB | PostgreSQL | Distributed SQL over the pg wire (YSQL, port 5433). |
| Amazon Aurora PostgreSQL | PostgreSQL | AWS-managed pg-compatible edition. IAM sign-in supported. |
| Amazon RDS for PostgreSQL | PostgreSQL | AWS-managed PostgreSQL. IAM sign-in supported. |
| Azure Database for PostgreSQL | PostgreSQL | Flexible server. TLS required. |
| Google Cloud SQL for PostgreSQL | PostgreSQL | GCP-managed PostgreSQL. IAM sign-in supported. |
| Neon | PostgreSQL | Serverless PostgreSQL. TLS required. |
| Supabase | PostgreSQL | Supabase-hosted PostgreSQL. |
MySQL wire
| Engine | Compatible with | Notes |
|---|---|---|
| MySQL | - | First-party. `backtick` quoting, LIMIT, ON DUPLICATE KEY upsert. |
| MariaDB | MySQL | Adds sequences and RETURNING; its own plan (ANALYZE FORMAT=JSON) and activity dialects. |
| Amazon Aurora MySQL | MySQL | AWS-managed MySQL-compatible edition. |
| Amazon RDS for MySQL | MySQL | AWS-managed MySQL. |
| Azure Database for MySQL | MySQL | Flexible server. TLS required. |
| Google Cloud SQL for MySQL | MySQL | GCP-managed MySQL. |
| PlanetScale | MySQL | Vitess-based. TLS required. No Maintenance menu or Security editor. |
| TiDB | MySQL | Distributed SQL (port 4000); no stored procedures, no OPTIMIZE TABLE. |
In-process file
| Engine | SQL dialect | Notes |
|---|---|---|
| SQLite | SQLite | File or in-memory database, opened directly. Dates and UUIDs are text-encoded; no stored procedures. Double-click a .sqlite, .db, or .sqlite3 file to open it. |
| DuckDB | PostgreSQL-family | Bundled - nothing to install. A .duckdb file or :memory:, no credentials. Real plans (EXPLAIN / EXPLAIN ANALYZE), read_parquet / read_csv / read_json offline. Connecting |
HTTP
| Engine | SQL dialect | Notes |
|---|---|---|
| ClickHouse | ClickHouse (MySQL-family grammar) | Its HTTP interface - port 8123, or 8443 with TLS. Results stream as ClickHouse sends them; wide integers and decimals keep their digits. No transactions, so the rollback wrap is refused up front. Connecting |
| libSQL / Turso | SQLite | Hrana over HTTP - a self-hosted sqld on port 8080, or Turso Cloud over HTTPS. Signs in with an auth token kept in your OS keychain. Connecting |
Oracle Net
| Engine | SQL dialect | Notes |
|---|---|---|
| Oracle Database | Oracle (FETCH FIRST, MERGE, PL/SQL) | Port 1521; service name, or SID: to connect by SID. Needs the free Oracle Instant Client on your machine - Test Connection tells you exactly what to install if it is missing. Connecting |
RESP - Redis family
| Engine | Compatible with | Notes |
|---|---|---|
| Redis | - | Not SQL: a query tab runs one command per line in redis-cli syntax, and the explorer is a key browser grouped by prefix. Port 6379, redis:// or rediss://. Connecting |
| Valkey | Redis | The Linux Foundation fork; wire-compatible. |
| KeyDB | Redis | Multithreaded fork; no FUNCTION LIST. |
| DragonflyDB | Redis | Redis-compatible in-memory store; CLIENT LIST omits the user column. |
| Amazon ElastiCache | Redis | Redis or Valkey; in-transit encryption means TLS Require or Verify full. Cluster mode is supported through the configuration endpoint (Redis Cluster). |
| Amazon MemoryDB | Redis | TLS required; the ACL user name goes in the Username field. Always a cluster: commands are routed by slot (Redis Cluster). |
What works where
Every tool that has a counterpart on an engine works against that engine's own catalog and system views. Where an engine has no counterpart, the menu item stays visible and says why - “DuckDB runs inside SQLly with no server sessions to monitor” - instead of failing after you click. Wire-compatible labels inherit their base engine's column below, except where their own row notes otherwise.
| Tool | SQL Server | PostgreSQL | MySQL / MariaDB | SQLite | libSQL / Turso | DuckDB | ClickHouse | Oracle | Redis |
|---|---|---|---|---|---|---|---|---|---|
| Explorer, IntelliSense, hover, Object Browser | Yes | Yes | Yes | Yes | Yes | Yes | Yes | Yes | Key browser |
| Estimated plan | Yes | Yes | Yes | Yes | Yes | Yes | Yes | Yes | - |
| Actual plan | Yes | Yes | Yes | - | - | Yes | - | - | - |
| Row editor | Yes | Yes | Yes | Yes | - | Yes | - | - | - |
| Maintenance menu | Yes | Yes | Yes | Yes | Yes | Yes | Yes | Yes | - |
| Server Dashboard | Yes | Yes | Yes | Yes | Yes | Yes | Yes | Yes | Yes |
| Activity Monitor | Yes | Yes | Yes | - | - | - | Yes | Yes | Yes |
| Query Insights | Yes | Yes | Yes | - | - | - | Yes | - | Yes |
| Disk Usage | Yes | Yes | Yes | Yes | Yes | Yes | Yes | Yes | Yes |
| Index Analyzer | Yes | Yes | Yes | - | - | - | - | - | - |
| Backup / Restore | Yes | Yes | Yes | - | - | - | Yes | - | - |
| Schema Compare & History | Yes | Yes | Yes | Yes | Yes | Yes | Yes | - | - |
| Data Compare | Yes | Yes | Yes | Yes | Yes | Yes | Yes | - | - |
| Data Transfer (as target) | Yes | Yes | Yes | Yes | - | Yes | Yes | - | - |
| Structure Editor | Yes | Yes | Yes | Yes | Yes | Yes | Yes | - | - |
| Test Data Generator | Yes | Yes | Yes | Yes | Yes | Yes | Yes | - | - |
| Dependencies (View Dependencies / Find Usages) | Yes | Yes | Yes | Yes | Yes | Yes | Yes | - | - |
| Security editor | Yes | Yes | Yes | - | - | - | Read-only view | - | - |
| Designers, Extended Events, SQL Agent | Yes | - | - | - | - | - | - | - | - |
| Procedure Mode | Yes | Yes | - | - | - | - | - | - | - |
| Manual commit mode | Yes | Yes | Yes | - | - | - | - | - | - |
| SSH tunnel, proxy, Kubernetes port-forward | Yes | Yes | Yes | - | Yes | - | Yes | - | Yes |
The reasons behind each dash are the app's own words. SQLite, libSQL and DuckDB keep no index usage counters, so there is nothing to analyze; DuckDB runs inside SQLly with no server sessions to monitor, and libSQL serves each request on its own with no session list; ClickHouse has no EXPLAIN ANALYZE and no row-level UPDATE (use ALTER TABLE … UPDATE in a query tab); Oracle plans come from EXPLAIN PLAN without executing, its backups run through RMAN or Data Pump on the server, and the Oracle client dials the listener directly, so tunnels and proxies do not apply yet; Redis has no schema, tables, indexes or execution plans, and manages users with ACL SETUSER. SQL Agent, Extended Events and the guided designers are SQL Server features and say so.
Connecting to the newer engines
SQL Server, PostgreSQL, MySQL and SQLite connect the way you would expect. The engines below each have one thing worth knowing before the first connection.
DuckDB
DuckDB is bundled with SQLly - there is nothing to install. Pick DuckDB
in the connection editor and enter a file path in the Server field (a new file is
created if it does not exist) or :memory: for a scratch database that
lives as long as SQLly runs. No credentials, no port. Double-clicking or dropping a
.duckdb or .ddb file opens it directly, and the connection
URL box accepts duckdb:/path/to/analytics.duckdb,
duckdb::memory:, or a bare .duckdb path.
DuckDB's lists, structs and maps appear as JSON text in the grid. Backups are the file itself, copy it, or run EXPORT DATABASE - and there is no Create Database dialog because a DuckDB database is a file. The Maintenance menu offers VACUUM, VACUUM ANALYZE, ANALYZE, CHECKPOINT and FORCE CHECKPOINT.
ClickHouse
Pick ClickHouse, enter the host, the HTTP port (8123,
or 8443 with TLS), a user and password, and optionally a database -
or paste a clickhouse://user:password@host:8123/database URL. Any
TLS mode other than Disable
switches the connection to HTTPS, and SSH tunnels, proxies and Kubernetes
port-forwards work as they do for PostgreSQL and MySQL.
The optional Native port field is the server's native-protocol TCP port. SQLly itself only talks HTTP; the field is used by Open in Terminal, because clickhouse-client speaks the native protocol. Leave it blank for the usual mapping (9000, or 9440 with TLS). On an SSH-tunneled connection the terminal gets its own forward to this port.
Scripts with several statements run one at a time and report each result set or the rows written; Cancel stops the query on the server. Query Insights reads system.query_log (top queries and recent failures), the estimated plan shows the parts and granules each index pruned, Data Transfer creates target tables as MergeTree, Backup / Restore scripts BACKUP / RESTORE DATABASE to the server's backups disk, and the Maintenance menu's one verb is OPTIMIZE TABLE … FINAL. Create and drop databases with CREATE DATABASE / DROP DATABASE in a query tab.
Oracle Database
SQLly's Oracle driver loads Oracle's free Instant Client (Basic or Basic Light) at connect time, so install that first from oracle.com and restart SQLly. If it is missing, Test Connection says exactly which library it looked for and where:
- macOS - the Apple Silicon (ARM64) DMG installer; SQLly looks for
libclntsh.dylibin~/lib,/usr/local/lib, orDYLD_LIBRARY_PATH. - Linux - unzip the Basic package and add its folder to
LD_LIBRARY_PATHor anldconfigdirectory; SQLly looks forlibclntsh.so. - Windows - unzip the Basic package and put its folder on
PATH; SQLly looks foroci.dll. - On any platform,
SQLLY_ORACLE_CLIENT_DIRset to the Instant Client folder overrides the search.
Then pick Oracle Database, enter the host, port (1521),
the service name in the Database field, prefix it with
SID: to connect by SID instead, or put a full connect descriptor in the
Server field for TCPS or wallets - and your username and password. Add
AS SYSDBA or AS SYSOPER after the username for an
administrative session.
PL/SQL blocks end with / on their own line, DBMS_OUTPUT lines appear in the Messages pane, and the explorer shows schemas, tables, views, packages, sequences, synonyms and types. The Maintenance menu gathers statistics (DBMS_STATS), validates structure and rebuilds indexes. Row editing, the Security editor, backup, and the compare and transfer tools are not available on Oracle yet, and each says so.
libSQL and Turso
Pick libSQL / Turso, enter the host, a …turso.io
address, or the machine running sqld on port 8080 - and
paste the database's auth token in place of a user name and password
(leave it blank for a local server without authentication). For Turso, create the
token with turso db tokens create <database> or from the Turso
dashboard; SQLly keeps it in your operating system's secure store, and a Turso host
is always reached over HTTPS. The URL box accepts
libsql://name-org.turso.io?authToken=…, or
http://127.0.0.1:8080 for a local sqld.
It is SQLite's dialect over the network, so the tools that read sqlite_master - Schema Compare, Data Compare, the Test Data Generator, the structure editor, dependencies - work as they do on SQLite. Scripts run as one unit, so BEGIN … COMMIT and the rollback wrap work; Cancel abandons the request rather than interrupting the server, and a whole result arrives before its first row, so page very large reads with LIMIT/OFFSET. The server refuses VACUUM and ANALYZE, so the Maintenance menu offers REINDEX and the integrity check. Row editing and Data Transfer into libSQL are not available yet.
Redis, Valkey, KeyDB, DragonflyDB, ElastiCache and MemoryDB
Pick Redis (or the compatible label that matches your server), enter
the host, port (6379), an ACL user name if the server uses one, the
password, an optional logical database number and a TLS mode, or paste a
redis://user:password@host:6379/2 or rediss:// URL.
Tunnels, proxies and port-forwards work as they do for the SQL engines.
- Commands, not SQL. A query tab runs one command per line in the same syntax as
redis-cli(quotes and escapes included, or a JSON list of arguments for unusual values). Every reply lands as a result set - hashes as field/value rows, sorted sets as member/score,INFOas section/key/value,CLIENT LISTandSLOWLOGas tables.SELECTswitches logical database for the rest of the script, and commands that would hang forever (SUBSCRIBE,MONITOR, a blocking pop with no timeout) are refused before they are sent. - Key browser. The explorer lists numbered databases with key counts and walks keys with a bounded, non-blocking scan, folding them into folders by the delimiter you chose (
:by default, souser:42:profilesits underuser→42). Each key shows its type and remaining TTL; double-click one to run the right reader for its type -GET,HGETALL,LRANGE,SMEMBERS,ZRANGEwith scores,XRANGE,JSON.GET- and the key menu offers Copy Key, Copy Command, Set TTL, Remove TTL and Delete Key, which confirm first. - Read-only means read-only commands. The SELECT-only lock and the AI's read-only mode allow only read-only commands; destructive ones -
FLUSHALL,FLUSHDB,DEL,CONFIG SET,SHUTDOWNand friends - go through the usual confirmation. Server Dashboard readsINFO, Activity Monitor readsCLIENT LIST(with kill), Query Insights reads the slow log, and Disk Usage shows memory per logical database.
Modeled - drivers on the way
These engines are fully described in the capability model, so SQLly can generate their SQL, quoting, and type mappings. They do not yet have native connectors, which means a live connection from the app is not available today.
| Engine | Connects via | SQL dialect |
|---|---|---|
| IBM Db2 | DRDA | Oracle-family |
| IBM Informix | DRDA | Oracle-family (SELECT FIRST n) |
| Microsoft Access | ODBC / ACE | SQL Server-family (reduced surface) |
| Snowflake | HTTP API | PostgreSQL-family (QUALIFY, MERGE) |
| Databricks SQL | HTTP API | PostgreSQL-family (Spark SQL) |
| Google Cloud Spanner | HTTP API | PostgreSQL-family (GoogleSQL, INSERT OR UPDATE) |
| Teradata | Proprietary | Oracle-family (SELECT TOP n, QUALIFY) |
| SAP HANA | Proprietary | Oracle-family |
One intent, per-engine SQL
Because engine rules live in the model, one UI action can render the right dialect wherever you run it. Row limiting is a familiar example: “view top 1000 rows” becomes:
| Pattern | Rendered SQL | Engines |
|---|---|---|
| TOP | SELECT TOP 1000 … | SQL Server, Azure SQL, Access, Teradata |
| LIMIT | … LIMIT 1000 | PostgreSQL & MySQL lineages, SQLite, libSQL, DuckDB, ClickHouse, Snowflake |
| FETCH FIRST | … FETCH FIRST 1000 ROWS ONLY | Oracle, Db2 |
| FIRST | SELECT FIRST 1000 … | Informix |
The same per-engine handling covers identifier quoting, string, boolean, and binary literals, upserts (ON CONFLICT, ON DUPLICATE KEY, or MERGE), and how all 15 logical types map to native storage. Redis, having no rows, maps every logical type to a binary-safe string and every write to a key set.