Engines

Six Engines, First-Class

Every engine gets its own executor, its own shell dialect and its own quirks handled — not a lowest-common-denominator SQL box. This page is the per-engine reference.

MongoDB

• Default port: 27017 · Auth: none, SCRAM-SHA-256, SCRAM-SHA-1
• The real shell experience: mongo-shell JavaScript with cursors, ObjectId/ISODate literals, runCommand, and multi-statement scripts
• Legacy servers 3.6 and 4.0 work out of the box — Sutido ships a dual driver and negotiates the right one automatically; no separate tool for that one old cluster
• Replica sets: multi-host lists, replica set name, read preference
• Visual Query Builder for find queries
• MongoDB Atlas via mongodb+srv:// — see Hosted Databases

PostgreSQL

• Default port: 5432 · Auth: username/password
• Multi-statement SQL scripts run on a single pinned session — SET, transactions and temp tables behave the way you expect across statements
• Abort cancels the query server-side via pg_cancel_backend
• The sidebar browses tables, views, materialized views, functions, procedures, triggers and sequences

Microsoft SQL Server

• Default port: 1433 · Auth: SQL logins or Windows authentication
• T-SQL scripts with proper GO batch handling — batches are sent whole, the way SSMS does it
• Configurable encryption and trust-server-certificate options
• Abort cancels the running request server-side

Aerospike

• Default port: 3000 · Auth: optional username/password
• An AQL shell: SELECT, INSERT, UPDATE, DELETE, index DDL, SHOW, STAT
• Namespace and set browsing in the sidebar
• Record metadata as real columns — _ttl (seconds until the record expires, -1 for never) and _gen, on scans as well as primary-key lookups
• Deletes are recoverable. AQL deletes by primary key, so the record is read and journaled before it goes; restore it from the changelog with its bins intact, maps and lists included
• Filtering a bin without a secondary index tells you so — Aerospike can only filter on an indexed bin, and the message carries the CREATE INDEX that would fix it instead of a status code
• An unfiltered SELECT says it read the whole set, because on a production namespace that matters
• The native driver runs isolated in its own process — a driver crash never takes the app down

ClickHouse

• Default port: 8123 (HTTP) · Auth: username/password
• Native ClickHouse SQL over HTTP — DESCRIBE, PREWHERE, aggregations over millions of rows
• Abort issues KILL QUERY server-side — runaway scans actually stop
• Sensible default row limits keep SELECT * on huge tables browsable
• Schema via DESCRIBE TABLE, right from the sidebar's View Schema action

Elasticsearch

• Default port: 9200 · Auth: none or username/password
• An index is addressed by name alone, so this engine has two levels where the others have three — the sidebar lists indices directly under the connection, with no database row in between
• Kibana Console syntax: a method and a path on one line, an optional JSON body under it. That is what an Elasticsearch user already types, and it puts both halves of the decision on screen — GET /orders/_search and POST /orders/_delete_by_query are the same shape of request to the same index, and one of them empties it
• Field names come from the mapping, not from a sample of documents, so the list is complete rather than however deep the sampler happened to look
• Deleting a single document is recoverable — DELETE /orders/_doc/1 is read back first and restored under the same id. A _delete_by_query is not, and the editor warns before running one
• Export walks the whole index with a scroll; import writes it back in _bulk batches, ids intact
• Plain HTTP for now. A connection cannot carry a TLS scheme yet, so a https:// URL is refused rather than quietly downgraded to cleartext with your password in the header

Next Steps

See Connection URIs for each engine's string format, and Authentication & Secrets for credential handling.