PostgreSQL
Chat with your PostgreSQL database
Point ChatPlotDB at a Postgres connection and ask questions the way you'd ask a colleague. It plans the query, writes the SQL, runs it read-only, and hands back a table or a chart.
ChatPlotDB is an AI data analyst for PostgreSQL. You register a connection with a host, port, database, user and password, pick the SSL mode your server requires, and start asking. Queries execute read-only against whatever role you connect with, and the SQL that produced each answer is shown beside it.
The questions people actuallybring to the data team
Anything you could answer with a SELECT, asked in English instead. No SQL to write, and no ticket to file.
“How many accounts signed up last month, broken down by plan?”
“Which customers haven't logged in for the last 30 days?”
“Show monthly recurring revenue by month for the past year”
“What's the average time between signup and first paid invoice?”
“Which ten accounts generated the most support tickets this quarter?”
“Compare trial-to-paid conversion this quarter against last”
Built for a real Postgres,not a demo schema
A production database has hundreds of tables, ambiguous column names, and joins nobody remembers. That's the case this is designed for.
Read-only by construction
ChatPlotDB only ever issues read queries. Point it at a role with SELECT on the tables you want queried and nothing more, and the blast radius is zero by definition.
The SQL is always visible
Every answer comes with the query that produced it. Anyone who reads SQL can check the logic; anyone who doesn't can ignore it entirely.
Encrypted credentials
Connection details are encrypted before they're stored and are never sent back to the browser. The interface shows a redacted summary, never the password.
SSL modes you choose
require, prefer, allow, or disable: pick the mode your server enforces when you register the connection.
Joins it works out itself
A config layer records which tables matter and how they relate, so a question spanning four tables becomes one correct query rather than four guesses.
Several databases at once
A stitched config composes between two and five databases into a single conversation, so one question can reach across your Postgres and the systems beside it.
Give it a read-only roleand nothing else
The safest posture isn't a promise in a marketing page: it's a Postgres role that cannot do the thing you're worried about.
ChatPlotDB issues read queries only. Even so, the right way to connect it is a dedicated database user with SELECT granted on the tables you want queried and no other privilege. If the credential can't write, no bug, prompt, or misread question can write either.
Credentials are encrypted before they are stored and are never returned to the browser: the connection screen shows a redacted summary. You choose the SSL mode your server requires when you register the connection: require encrypts without verifying the certificate, prefer encrypts when the server supports it, allow encrypts only when the server insists, and disable turns it off. Certificate-verifying modes exist in the data model but are not selectable yet, because there is nowhere to supply a CA certificate.
After a migration changes your schema, re-run Config Studio so the config reflects the new tables. The agent plans against that config rather than reading your live schema on every question, which is what keeps the answers consistent, and what means a schema change is something you tell it about once.
Common questions
No. It only ever issues read queries. The recommended setup is a dedicated Postgres role with SELECT granted on the tables you want queried and nothing else, so the credential itself cannot modify anything.
require, prefer, allow and disable. You choose the mode when you register the connection, matching whatever your server enforces.
verify-ca and verify-full exist in the data model but are not selectable yet: they need a CA certificate, and there is nowhere to supply one.
Encrypted before they are written, and never sent back to the browser. The connection screen shows a redacted summary rather than the values you typed.
Yes. A stitched config composes between two and five databases into a single conversation, so a question can reach across them without you exporting anything first.
Re-run Config Studio so the config picks up the new tables. The agent plans from the config rather than the live schema, so it keeps answering consistently until you tell it the shape has changed.
There is also a Test action on every connection that re-opens it and reports failures without changing anything.
Always. The generated query is shown alongside the answer, so a colleague who reads SQL can confirm the logic, which is usually how a team builds confidence in the config in the first place.
