PostgreSQL family
PostgreSQL, CockroachDB, Redshift
Point Ottermind at the database you already query. Ask in plain language and get a traceable HTML report, a slide deck, or a weekly analysis that keeps running - with the SQL it ran shown next to every number.
First step
Analysis runs against a source you save once. Here is exactly what Ottermind asks for, what it is allowed to do with it, and where the query actually runs.
PostgreSQL family
PostgreSQL, CockroachDB, Redshift
MySQL family
MySQL, MariaDB, TiDB, OceanBase
Warehouses
ClickHouse, Snowflake
Enterprise
SQL Server, Oracle
Local and embedded
SQLite, DuckDB
No database at hand? A local SQLite or DuckDB file works as a source, and so do spreadsheets, CSVs and documents - useful for trying the workflow before you involve a production database.
Coming from Chat2DB?
Chat2DB is a local-first SQL workspace - 40+ engines, schema browsing, editing, execution. It is where you make sure the number is correct. Ottermind picks up at the next step: when that number has to become a report your team reads, a deck for Monday, or an analysis that runs again next week across more than one source. Same databases, same team.
Chat2DB - open-source SQL workspaceNo modelling layer to configure and no semantic layer to maintain. Connect the database, describe the question, keep the artifact.
Step 1
Add PostgreSQL, MySQL, SQL Server, Oracle, ClickHouse, Snowflake, SQLite or DuckDB under Settings - Data sources. Credentials are stored encrypted and are never passed through the conversation.
Step 2
Describe the question, not the schema. Ottermind inspects the tables it needs, writes the SQL, and runs it inside a read-only transaction with a bounded execution budget.
Step 3
A self-contained HTML report, a slide deck, or a spreadsheet - with every figure traceable to the query that produced it.
Both are built by the same team. They are not built for the same job.
| SQL client (Chat2DB) | Ottermind | |
|---|---|---|
| Unit of work | One query | One deliverable |
| Who it is for | Developers and DBAs | Analysts, operators, founders |
| You end up with | A result set, a chart, a dashboard | A report, a deck, a written conclusion |
| Engine coverage | 40+ engines over JDBC | 8 engine families, cloud or your own desktop |
| Across sources | Inside one database | Several databases, plus files and documents |
| Time span | This query, right now | A question you re-run every week |
Engine coverage is where a dedicated SQL client wins, and we would rather say so than pretend otherwise: if you need Hive, DB2, Elasticsearch or Trino, query it in Chat2DB.
The point is not the chat transcript. It is the artifact at the end of it.
Headings, accessible tables, legible charts, units and metric definitions - in a single file with no external scripts or trackers, linked back to the query behind it.
The same analysis as a visual deck: the trend, the breakdown, the two things that actually moved, and what you are recommending.
Run the same analysis again on a schedule and get told what changed since last time, instead of rebuilding the query from memory.
Join what the database says with a spreadsheet, a PDF contract or a document in your drive, for the times the answer is not all in SQL.
An agent with database credentials should be boring and predictable. These are the limits it runs under.
Analysis runs SELECT and WITH only. Write statements are rejected, including ones smuggled into a query argument.
Single-table inserts, updates and deletes require the source's explicit operation permission plus your confirmation on the prepared change. Supported on PostgreSQL, MySQL, MariaDB and SQLite.
Every figure can be traced to the statement that produced it. A result page is a page, not the whole dataset, and the report is required to say so.
Passwords are stored encrypted and handed to the executing device, never placed into the chat context or written to a file.
No. Chat2DB is a SQL workspace: connect, browse, edit, run, and be certain the number is right. Ottermind is what happens after that - writing the analysis up, turning it into a deck, and re-running it next week. Most people who use both keep querying in Chat2DB.
PostgreSQL, CockroachDB and Redshift; MySQL, MariaDB, TiDB and OceanBase; SQL Server; Oracle; ClickHouse; Snowflake; SQLite and DuckDB. If your engine is not on that list, a dedicated SQL client with JDBC coverage is the better tool for it.
Yes, through the Ottermind desktop app bound to your account. The desktop executes the query locally, so a database that only listens inside your network does not need a public endpoint. Cloud execution is used only for sources that are reachable from the internet.
Not during analysis. Queries are restricted to SELECT and WITH inside a read-only transaction. Changing rows is a separate, permissioned path: the source has to grant write operations, the change is prepared first, and you confirm that specific change before anything is committed.
No. Describe the question in plain language and Ottermind inspects the schema and writes the SQL. It still shows you the statement it ran, which is the part that matters when someone asks where a number came from.
A single self-contained HTML file: headings, tables, charts drawn as SVG, stated units and metric definitions. It loads no external scripts, fonts or images, so it behaves the same when you send it to somebody else.
Yes. The same analysis can be delivered as a visual slide deck rather than a report, which is usually what a quarterly review or a board update actually needs.
Spreadsheets, CSVs, PDFs and documents work as sources too, and can be combined with a database in the same analysis. A local SQLite or DuckDB file counts as a database source.
Aggregates are computed in SQL rather than read off a page of sample rows, time windows and metric definitions are stated, and sampling or missing data has to be disclosed in the deliverable. Paging through results re-runs the query, so a re-opened page reflects the source as it is now, not a frozen snapshot.
Data analysis is part of the Ottermind workspace rather than a separate product. See the pricing page for current plans.
Connect a database and ask for the analysis you would otherwise spend the afternoon assembling by hand.