The word implies a second pair of hands on the controls. This one deliberately has none — it reads, suggests and explains, and every write stays yours.
A new engineer is productive in your codebase in a week and in your database in a quarter. The database is where the undocumented history lives: the table that was renamed but not migrated, the boolean that means three things, the join everyone knows to add. A copilot that has read the whole catalog — keys, types, distinct values, statistics — collapses a lot of that into asking a question and reading the query that comes back.
Coding copilots are safe because their output lands in a diff you review before anything runs. A database copilot has no diff — the statement executes against live data and there is no undo. Removing writes entirely restores the property that made the coding case work: everything it does is reversible, so you can let it act rather than approving each step. Generated SQL is parsed to an AST and validated to a single read-only SELECT; the guarantee is structural rather than a matter of prompting.
When the answer to "why is this slow" is an index, you get the `CREATE INDEX` statement to read, review and apply through whatever migration process you already have. That is the correct boundary. An index is a permanent write-path cost on a table, and the judgement about your workload belongs to someone who knows what that table is for.
The desktop app is per-engineer and needs nothing central. For a shared deployment there is a server build: one instance, read-only credentials supported throughout, and every answer recording the query it ran. If that is the shape you need, talk to us — it is a conversation rather than a checkout.
More answers on the main FAQ, or ask us directly.
For people who already write SQL
The category, explained
Evaluation
Every page in this section covers a different question about the same product. See all solutions.
Start free, upgrade when you are asking enough questions for it to matter.
