🚀 launched today: 🐘 Postgres branches for every AI agent. → build a Postgres branch in ~3 seconds → production-like data → isolated database for every agent → sensitive data automatically masked → branch → work → delete Already ~100 Postgres branches/week before public launch. Now we're building the... Postgres runtime for agents.show more

$10 free credit to try it. REPLY with your PG version!

Postgres 18

does it work with planter scale postgres?

Yes 👍, with any cloud.

Postgres 18

write DM!

Every agent gets its own Postgres. No shared staging. No conflicts. No waiting.

You are driving at another speed solving real problems. Man, you are crazy. Love 🥰 this vibe. Thank you 🙏

What is your postgres version)?

Hmmm. 🤔 got me. 18.2? Or 18.5?

Tnx, will give some creds. dm

Thanks, appreciated but needed. We needed to stay compatible with IBM cognos. Upgrade to IBM 12.1.3 FP 2 allowed us to move. Had no time to test you things yet. It’s on my todo list, way up 🤣

No worries at all 😄 Glad the upgrade finally unblocked you! Whenever you get a chance to try it, I’d love to hear your feedback.

Isolated branch plus masked prod-like data is the right seat — agents get a database without writing the live one.

The branch, work, delete loop is the right primitive for agent experimentation. Isolation makes retries cheap without turning the database into a shared-state mystery. Are agents creating their own branches reliably yet?

Yes agents can already create, use, and delete their own isolated Postgres branches reliably.

Looks really cool

Tnz for feedback, try it)

I love it, it’s the same instinct that led me to create Fottly as a self-hosted solution instead of opting for another managed image API. Having control over the infrastructure of something you use constantly makes perfect sense.

What is your DB version?

No DB on the self-hosted side, it's stateless, images go straight to S3-compatible storage. Which means right now I have zero visibility into who's actually running it. Your question just made me realize I should add anonymous opt-in telemetry to fix that. Thanks!

you can try pgrun!

branch per agent in 3 seconds is nuts, the auto-masking bit is what makes this feel production-shaped

The killer feature isn’t the 3-second branch. It’s the delete button - clean rollback beats debugging shared state after five agents touched it.

branching with production-like data in 3 seconds changes what you can safely let an agent touch

Yeees, 3-5 seconds

Looks great, are you planning to open source it?

sure, will do soon!

This is cool! Does it handle seeding as well or just the branches?

Yep! It handles seeding too branches can start with production-like data, automatically masked.

Dope! We’re on Planetscale so I’ll def have to check this out

Very cool! Branches are the main reason I use neon. This seems similar? With neon, when I run the whole integration suite it creates ~500 branches in parallel in a few seconds, runs the tests, and is done in about 10-20 min. I don't really know how many times per day agents do it, but they do it all day long and it costs me around $20/month. How do you compare?

Your production database stays where it is. Connect your existing Postgres once. Then agents, CI, and pull requests can create isolated, disposable branches on demand.

3 seconds is insanely fast

Even if you have 200GB db!

pg16)

you have to update it!)

If data is masked per branch, how does the system handle referential integrity when an agent needs to join records across two isolated tables? Linking based on a partial key might fail if the raw value was necessary for the foreign key constraint check.

Good question