ClickHouse Acquires Xata and Expands Into Postgres, AI Tooling

ClickHouse has acquired Xata and expanded its platform into managed Postgres, observability, security and AI engineering, a move the company says will consolidate systems that businesses currently run separately as AI agent adoption accelerates.

The announcement, made on November 10, 2026, adds four areas to the core ClickHouse platform. Transactional database services arrive through ClickHouse Managed Postgres. Observability and security products join them, drawing on technology from RunReveal, a company ClickHouse acquired earlier. The AI engineering element is built around Langfuse, the open-source platform ClickHouse had already purchased, which now ships with an AI Gateway designed to give organisations more oversight of how AI models are used across applications.

The Xata acquisition is intended to strengthen the managed Postgres offering. The deal will let developers branch and clone databases quickly, creating isolated Postgres environments for development and testing, while also supporting work on a separate tier aimed at AI agents.

Why does one platform matter for AI workloads?

The core argument ClickHouse makes is about workload behaviour. AI agents can create uneven and unpredictable patterns of database activity as they query and test multiple systems. Running those workloads through several separate tools raises cost and operational complexity, according to the company. ClickHouse is positioning its platform as a way to handle analytics, transactions, monitoring and security in one environment, a direct challenge to established suppliers that serve those categories individually.

That strategy rests on a single underlying system serving functions that typically sit in different vendor stacks. The company argues that as AI workloads spread, the fragmentation across analytics, transactions, monitoring and security becomes harder to justify operationally. Consolidation on one foundation is the pitch.

The release stages vary:

• ClickHouse Managed Postgres — generally available now in ClickHouse Cloud
• ClickHouse Observability — public beta
• ClickHouse Security — private preview
• Langfuse AI Gateway — public beta

Further additions tied to the Xata deal have not yet shipped. Those include instant database branching, cloning and scale-to-zero environments. ClickHouse is also laying the groundwork for a dedicated product tier aimed specifically at AI agents.

Who is using the platform?

ClickHouse now serves more than 5,000 customers. It counts DoorDash, Ramp, Meta, Tesla, Cisco and Visa among the organisations using its platform for data-intensive workloads.

Mercado Libre, which ClickHouse described as Latin America's leading commerce and fintech platform, is one customer already running more than one workload on the system. The company uses ClickHouse for observability and processes 400 million spans per minute on it. ClickHouse said Mercado Libre has seen query speeds improve by 75 times and compression reach roughly 89%, reducing infrastructure costs.

That figure gives the consolidation pitch its most concrete comparison: a single observability pipeline handling 400 million spans per minute while compressing storage to about a ninth of its raw size.

Physical Intelligence, a company working on general-purpose AI for robotics, is another customer expanding across the platform. One of its engineers described the appeal of building on a single foundation across multiple tasks.

"Building robotics foundation models means making sense of billions of annotations and telemetry records," said Karan Dhabalia, Engineer, Physical Intelligence.

"We started with ClickHouse for observability, then expanded into real-time analytics, Postgres and AI development with Langfuse. This shared foundation simplifies operations and helps our researchers develop better models. With ClickHouse Security now in private preview, we're excited to see the platform grow," said Dhabalia.

The Physical Intelligence deployment illustrates the adoption pattern ClickHouse is betting on: customers arrive for one workload, observability in that case, then extend into analytics, transactional Postgres and AI engineering tooling on the same foundation.

What changes for developers?

The Xata technology addresses a specific developer pain point. Branching and cloning databases quickly means teams can spin up isolated Postgres environments for development and testing without lengthy provisioning, and the planned scale-to-zero environments would let those copies cost nothing when idle.

For AI agent development, that isolation carries extra weight. Agents probing multiple systems generate database activity that is difficult to forecast, and isolated environments let developers test that behaviour without touching production data. The planned dedicated tier for AI agents would extend that separation to the commercial and operational layer of the platform.

The Langfuse AI Gateway, now in public beta, targets a different problem: oversight. As organisations deploy AI models across many applications, the gateway is intended to give them visibility into and control over how those models are used — a governance function that has typically required separate tooling.

How does this position ClickHouse against incumbents?

The expansion puts ClickHouse into competitive contact with vendors in four distinct categories at once: managed Postgres providers, observability platforms, security analytics suppliers and AI engineering toolmakers such as those around LLM observability.

ClickHouse's answer to that fragmentation is economic as much as technical. By combining the functions on the same underlying system, the company argues it can cut the cost and operational burden that comes from stitching together separate suppliers for each category. The Mercado Libre results — 75 times faster queries and roughly 89% compression — are the evidence it points to for what a single foundation can deliver at scale.

The sequencing matters for how quickly that argument can be tested. Managed Postgres is generally available today, so transactional workloads can move now. Observability and the AI Gateway are in public beta, open to customers willing to accept early-stage software. Security remains in private preview, limiting its immediate reach. The Xata-derived capabilities — instant branching, cloning and scale-to-zero — have no release date yet.

The customer base of more than 5,000, including DoorDash, Ramp, Meta, Tesla, Cisco and Visa, gives ClickHouse an installed audience for the new products. Whether those customers follow Physical Intelligence's path from a single workload to four will determine whether the consolidated platform thesis holds against incumbents defending each category separately.

The next milestones to watch are the general availability of ClickHouse Observability, the public release of ClickHouse Security, and the launch of the Xata branching and cloning features — each of which moves a piece of the consolidated platform from promise to product.