Built into Doozy
An AI coworker for Postgres, in Slack.
Connect a Postgres database and ask it questions in plain words. Doozy reads; it can't change a row.
- Works in
- Slack
- Connects by
- A connection string, stored encrypted
- Can do
- 5 things, each with a switch
- Price
- Included on every plan
What Doozy can do in Postgres
Give Doozy a read-only connection to your Postgres database and it answers questions from your real data: it finds the tables, reads their shape and runs the query, then posts the answer, a sheet or a chart in the thread. The connection refuses writes, so a question can't become a change.
- List the schemas and tables, with how many rows each holds
- Describe a table: columns, keys and indexes
- Run SELECT queries and return the rows as text, a sheet or a chart
- Follow notes you leave it about what the tables mean
- Run the same question on a schedule and post the result
Things people ask
In a thread, after @Doozy. It answers in that thread, with what it did.
@Doozy how many orders shipped late last month, by warehouse? Chart it.
@Doozy which customers signed up this week and haven't placed an order?
@Doozy what does the `subscriptions` table look like, and which column holds the plan?
@Doozy every Monday at 9, post last week's revenue by product.
How to connect Postgres
A few minutes, by whoever owns the workspace. No engineer needed.
- 1
Make a read-only database user, or use one you have.
- 2
On the Integrations page, choose Postgres and paste the connection string. Doozy checks it can reach the database.
- 3
Add a note on what the tables mean, if you like, and ask.
What stays off, and what's always true
Doozy works like a teammate, so it doesn't stop to ask. You decide what it can reach.
Off until you say so
- Writing to the database: never. The connection is read-only and each query stops after 30 seconds.
Always
- Every task runs in its own isolated workspace. Credentials stay outside it.
- Every tool has a switch. Tools that delete or overwrite start off.
- Every action it takes is on the audit log, with who asked and what happened.
- Owners and admins connect tools and choose switches. Members can only look.
More in Security and AI & Data Use.
Works well with
One ask often crosses tools: a Sentry error becomes a GitHub pull request and a Linear issue. Doozy uses whichever it needs.
Questions people ask before they connect Postgres
- Can it see the connection string later?
- No one can, Doozy included. It's stored encrypted and added to the connection outside the task's workspace, so the AI never holds it.
- What about MySQL or other databases?
- Postgres is the database Doozy connects to directly today. Supabase has its own connector, and other databases can be reached through an MCP server.
- Can Doozy see our Postgres credentials?
- No. Every task runs in its own isolated workspace, and credentials stay outside it: they're added to each call on its way out. Nothing it reads or writes there can expose a key.
- Will it change things in Postgres without asking?
- It works like a teammate, so it doesn't stop to ask, and you decide what it can reach. Tools that delete or overwrite start switched off, you can switch off any other, and every action it takes is on the audit log with who asked.
- Who can connect it?
- Owners and admins of your Doozy workspace. Members see what's connected and can't change it. Everyone in your Slack can ask Doozy once it's connected.
- What does it cost?
- Nothing extra. Every integration is included on every plan, and the 14-day trial needs no card.
Something else? Write to [email protected].
Your next hire is one mention away.
Add Doozy to Slack and hand it the thing that's been waiting on someone all week. It comes back done, and it starts learning how your team works from the first ask.
- 14 days free, no card
- One price for the whole team
- Nothing billed until you pick a plan