Postgres
by the thousands
The shift
Your infrastructure was built for humans
One developer, a few PRs, maybe 5 concurrent test environments. AI agents changed the math. An agent can open 50 PRs in parallel — each needing its own database. The infrastructure that scaled to 100 developers doesn't scale to 100,000 ephemeral copies.
| Human-led | Agent-scale | |
|---|---|---|
| Staging environments | 1-2 per team, shared | 1,000s per day, isolated |
| Lifespan | Weeks to months | Minutes to hours |
| Provisioning | Manual, hours | Automatic, seconds |
| Storage cost | $100s/mo × TBs × DBs | CoW = deltas only |
| Idle cost | $ per hour, always on | Scale-to-zero, $0 |
Platform
How it all fits together
Copy-on-Write Branching
Fork a live database in seconds. Branch from production or spawn from a snapshot — each copy shares storage with its parent and diverges independently.
Learn moreCopy-on-Write Storage
1,000 copies of a 100GB database don't cost 100TB. Storage is shared at the block layer via NVMe-over-Fabrics. You pay for deltas, not duplicates.
Learn moreScale-to-Zero Compute
Idle branches shut down automatically and restart on the next connection. Run thousands of branches — pay only for active queries.
Learn moreZero-Downtime Migrations
pgroll serves old and new schema versions in parallel. No table locks, instant rollbacks. Safe schema changes without deployment windows.
Learn more2-3s
Branch creation
60-90%
Cost reduction vs full replicas
$0
Idle branch compute cost
50,000+
Branches per storage node
Data & security
Production-safe from day one
Built-in Anonymization
From basic column masking to HIPAA-grade microaggregation. PII is detected automatically and anonymized during replication — every branch starts clean.
Learn moreBYOC Deployment
Deploy the data plane in your own AWS, GCP, Azure, or Hetzner account. Fully managed by Xata, data stays in your cloud. Use your negotiated discounts.
Learn moreHigh Performance
NVMe/TCP and SPDK technology for ultra-low latency. Separation of storage and compute. Smart tiering from hot NVMe to cold S3.
Learn moreAdoption
No production migration required
Connect to your existing RDS, Aurora, Cloud SQL, or self-hosted Postgres via logical replication. Production stays exactly where it is. Xata handles staging, dev branches, and anonymization on top.
100% vanilla Postgres
Not a fork. Every tool, ORM, and extension works unchanged. pg_dump and go if you leave.
Any cloud, any region
AWS, GCP, Azure, Hetzner. BYOC keeps data in your account. Use your cloud credits.
Open source core
pgroll, pgstream, Xata Agent — all Apache 2.0. No vendor lock-in, no source-available games.