Launching today
Chronos DB
Postgres branching and merge rules for AI agents
7 followers
Postgres branching and merge rules for AI agents
7 followers
The database for the autonomous AI era: every agent works in its own copy of your data, and changes reach main only by your rules or your approval. Postgres on the wire.
Free
Launch Team
Atlassian LoomHit record, hand off to AI
Promoted
Hi Product Hunt, I’m Abhishek, and I build Chronos with my co-founder Vaishak.
I started this while building the tech stack for a real estate company. I wanted an AI agent to update our real CRM data, and there was no safe way to do it. I could keep the agent read-only and apply its suggestions by hand, or give it the password and hope.
Chronos gives every agent its own copy of the data. When the agent finishes, your rules decide what happens. By default, up to 500 changed rows and no deletes merge on their own. Anything else waits for a person, who sees the change in plain words, like “Remove lead Hooli”, instead of a SQL log. If an approved change turns out wrong, you can undo that merge, row by row.
There are two ways to use it:
Starting something new: Chronos DB is a database built for agents. Each agent gets its own copy in about 2 µs, whatever the size, with search built in. It speaks Postgres.
Already on Postgres: Chronos Gateway runs next to the Postgres you already have. Agents get a token, not your password, and work over SQL or MCP. Every row is checked against live data as it’s written, so a stale agent can’t overwrite someone’s edit.
To be upfront about where it stands:
It’s self-hosted, and Chronos DB is in public beta for macOS and Linux.
The license is BUSL-1.1, so it isn’t open source.
Gateway is tested on Postgres 12 to 18, Supabase and Neon. RDS and Cloud SQL aren’t tested yet.
If your agents write to a database today, what’s the write you’d least want them to get wrong? I’d like to test Chronos against it.