Ardent gives engineering teams a way to clone any Postgres database in under six seconds, regardless of size, so coding agents and developers can run their work against a one-to-one copy of production or dev data with no risk to the live system. The product targets the testing gap that opens up when tools like Claude Code and Codex generate database changes that pass on a thin local seed but break against real schemas, real volumes, and real edge cases in production. Instead of waiting hours for a staging dump or settling for a sparse local fixture, an engineer runs `ardent branch create` and gets an isolated, fully writable Postgres branch backed by production data.
Under the hood, Ardent replicates the production database into a read replica and then creates lightweight copy-on-write forks on demand, with each fork functioning as a fully working Postgres instance . The workflow runs as a CLI modeled on git, with cloning as simple as ardent branch create , and connections to RDS Postgres instances require no config changes . The launch matters now because agent-driven coding has pushed database review into the bottleneck, and Ardent is already operating at meaningful scale with customers like Supermemory and Surface Labs running on 10TB-plus datasets.
Ardent was founded in 2025 by Vikram Chennai, previously an ML engineer and a Cornell CS alum , and is part of Y Combinator's Spring 2026 batch . The company is positioning database branching as core infrastructure for teams that want autonomous agents to ship migrations and queries without the 2 a.m. rollback cycle.
Comments (7)
Sign in to join the discussion.
N
Nikhil ParabMay 9, 2026
6 seconds to clone a TB? Either this is wizardry or someone is very loose with the word 'clone'. I want to see the copy-on-write tricks under the hood.
P
Pieter VandenbergMay 9, 2026
Every demo of 'safe agent on prod' I've seen quietly runs on a sanitized fixture. Curious what happens when an agent decides to DROP TABLE on the clone and the clone has a foreign key back to something it shouldn't.
Y
Yuki TanizakiMay 9, 2026
'Use coding agents on real data with no risk' is fine but a bit corporate. Try 'Let your agents touch prod. Nothing breaks.'
M
Marisol OteroMay 9, 2026
The launch video buries the 6s clone reveal at like the 18 second mark. That number IS your hook, lead with it and cut the intro card.
F
Folake EzeMay 9, 2026
Devrel question: what's the rate limit on clone creation per minute, and is there a webhook when a clone is ready or do I have to poll like it's 2014?
K
Krishan M.May 9, 2026
Reminds me a lot of what Neon and Supabase have been pushing on branching. The wedge here seems to be agent-shaped workflows specifically, which is a real one if the latency claim holds.
Z
zoltanMay 9, 2026
This might be the last DB tool launched where the founders actually wrote the migration scripts themselves. By Q2 the agents will be shipping the agents.
Work With Us
Book an intro call to explore our suite of services. We might have a waitlist.
Ardent × YC
Use coding agents on real data with no risk
@ArdentAI
Use coding agents on real data with no risk
Tags
TLVC Rating
Community Rating
No ratings yet
to rate this launch.
Ardent gives engineering teams a way to clone any Postgres database in under six seconds, regardless of size, so coding agents and developers can run their work against a one-to-one copy of production or dev data with no risk to the live system. The product targets the testing gap that opens up when tools like Claude Code and Codex generate database changes that pass on a thin local seed but break against real schemas, real volumes, and real edge cases in production. Instead of waiting hours for a staging dump or settling for a sparse local fixture, an engineer runs `ardent branch create` and gets an isolated, fully writable Postgres branch backed by production data. Under the hood, Ardent replicates the production database into a read replica and then creates lightweight copy-on-write forks on demand, with each fork functioning as a fully working Postgres instance . The workflow runs as a CLI modeled on git, with cloning as simple as ardent branch create , and connections to RDS Postgres instances require no config changes . The launch matters now because agent-driven coding has pushed database review into the bottleneck, and Ardent is already operating at meaningful scale with customers like Supermemory and Surface Labs running on 10TB-plus datasets. Ardent was founded in 2025 by Vikram Chennai, previously an ML engineer and a Cornell CS alum , and is part of Y Combinator's Spring 2026 batch . The company is positioning database branching as core infrastructure for teams that want autonomous agents to ship migrations and queries without the 2 a.m. rollback cycle.
Comments (7)
6 seconds to clone a TB? Either this is wizardry or someone is very loose with the word 'clone'. I want to see the copy-on-write tricks under the hood.
Every demo of 'safe agent on prod' I've seen quietly runs on a sanitized fixture. Curious what happens when an agent decides to DROP TABLE on the clone and the clone has a foreign key back to something it shouldn't.
'Use coding agents on real data with no risk' is fine but a bit corporate. Try 'Let your agents touch prod. Nothing breaks.'
The launch video buries the 6s clone reveal at like the 18 second mark. That number IS your hook, lead with it and cut the intro card.
Devrel question: what's the rate limit on clone creation per minute, and is there a webhook when a clone is ready or do I have to poll like it's 2014?
Reminds me a lot of what Neon and Supabase have been pushing on branching. The wedge here seems to be agent-shaped workflows specifically, which is a real one if the latency claim holds.
This might be the last DB tool launched where the founders actually wrote the migration scripts themselves. By Q2 the agents will be shipping the agents.