ClickHouse acquires Xata to bring branching and agentic features to Managed Postgres

Xata, the company behind a Postgres platform known for instant provisioning and copy-on-write database branching, is being acquired by ClickHouse. In an announcement published November 9, Xata said the two companies will build a combined Postgres platform aimed at modern applications and AI agents, pairing ClickHouse's high-performance production service with Xata's tooling for creating large numbers of short-lived databases. The post did not disclose financial terms.

What each side brings

According to Xata, the pairing matches two complementary services. ClickHouse Managed Postgres, which launched earlier this year and reached general availability on the day of the acquisition announcement, is built on local NVMe storage and is positioned by the company as the fastest Postgres offering on the market. It is tuned for high-throughput, long-lived production workloads.

Xata, by contrast, is designed to spin up many small, short-lived databases in seconds, populated with production-like data. Its headline features are instant provisioning, copy-on-write branching, scale-to-zero compute and anonymization of sensitive data before it leaves production. Both platforms run unmodified vanilla Postgres, which Xata says should make integration straightforward.

The bottleneck Xata targets

Xata's framing of the deal is that AI-driven development has shifted the bottleneck to the database layer. As coding agents and automated pipelines produce more output, every pull request and every agent run needs a safe, isolated database with realistic data, something traditional cloning cannot deliver at terabyte scale, where copying a production database is slow and expensive.

Copy-on-write branching avoids duplicating data outright: Xata says branching a multi-terabyte database takes roughly three seconds, and the child branch is fully isolated, with its own compute running on a separate physical node. Billing follows the diff rather than the copies, and Xata gives the example of fifty branches of a 1TB database being charged closer to 1TB than 50TB. Idle databases suspend automatically and wake in about a second.

Two technologies make this work. Xatastor, Xata's storage engine, separates storage from compute and exposes ZFS volumes over the network using NVMe/TCP; because it operates at the block-device level, ordinary PostgreSQL runs unmodified. Warm pools of pre-provisioned compute instances keep provisioning, wake-up and branching times short by taking node setup out of the critical path.

Postgres plus analytics

The acquisition also deepens a broader convergence play. Xata describes Postgres as the leading OLTP database and ClickHouse as the leading OLAP one, and notes that ClickHouse's Postgres team has been building first-class replication from Postgres into ClickHouse alongside a unified query layer, an integrated transactional and analytical stack that teams have historically stitched together themselves.

Xata lists several AI-native customers, including Sapien, Enginy, David AI, Runner AI and LoopAI, that it says create thousands of database branches per week. David AI, which already uses both platforms, said it moved from Neon to ClickHouse Managed Postgres for better reliability and performance, and that bringing branching and managed Postgres together would simplify its infrastructure. Xata, founded roughly five years ago, also maintains open-source Postgres tooling: pgstream, for logical replication with anonymization and DDL support, and pgroll, for reversible, zero-downtime schema changes.

What happens next

For now, nothing changes for Xata customers: the service keeps running and continues to accept new sign-ups. In parallel, the team is working on bringing branching and a dedicated agentic tier to ClickHouse Managed Postgres, and Xata customers will eventually be migrated onto that platform. ClickHouse published a parallel announcement of its own, though this article draws on Xata's post.

The anticipated use cases include an isolated database branch for every pull request via a GitHub integration that plugs into CI/CD pipelines, a branch for every agent run, and AI platforms embedding Postgres that must manage fleets of large numbers of mostly idle databases at low cost.

Why it matters

The deal is a notable consolidation in the database market and a signal of where infrastructure demand is heading. AI coding agents are turning isolated, production-like databases from a niche convenience into a baseline requirement, and the economics of copy-on-write branching plus scale-to-zero make it feasible to give every PR and every agent its own environment. At the same time, the acquisition pushes ClickHouse beyond its analytical roots into the transactional Postgres market, pairing OLTP and OLAP under one roof and sharpening competition with other Postgres platform providers. For engineering teams, the outcome to watch is whether branching and the agentic tier land in ClickHouse Managed Postgres quickly enough to keep Xata's existing customers, and the thousands of weekly branches they create, from feeling friction during the transition.