Observability teams are collecting more data than ever. The rise of AI applications, agentic workflows, and increasingly distributed systems has driven telemetry volumes to levels that would have been difficult to imagine only a few years ago. 

As a result, ClickHouse has increasingly become the database of choice for organizations looking to store and analyze observability data at scale, including [OpenAI](https://clickhouse.com/blog/why-openai-uses-clickhouse-for-petabyte-scale-observability), DoorDash, [Anthropic](https://clickhouse.com/blog/how-anthropic-is-using-clickhouse-to-scale-observability-for-ai-era), and Shopify, combining fast query performance with the cost efficiency needed for long-term retention. At the same time, Datadog has established itself as one of the industry's most popular observability platforms.

Historically, organizations often had to choose between retaining complete telemetry datasets and keeping that data accessible through the tools their teams use every day. Decisions about retention periods, sampling strategies, and storage locations were frequently driven by cost considerations rather than operational requirements.

Today, we're excited to announce a new partnership between ClickHouse and Datadog that helps remove those tradeoffs. With new capabilities now available in Preview, organizations can route logs directly to ClickHouse through Datadog Observability Pipelines and search those logs from the Datadog Log Explorer, combining ClickHouse's economics and performance with the Datadog experience.

## Route logs to ClickHouse with Datadog Observability Pipelines

Datadog Observability Pipelines provides a flexible way to collect, process, and route telemetry from a wide range of sources. Supporting open standards, such as OpenTelemetry and OCSF, it enables organizations to centrally manage telemetry while retaining control over where data is stored.

![datadog_to_clickhouse.png](/uploads/datadog_to_clickhouse_28cf8ed604.png)

With a native ClickHouse destination, logs can now be routed directly into ClickHouse through Observability Pipelines. Before data reaches ClickHouse, teams can leverage parsing, enrichment, filtering, transformation, and redaction capabilities to ensure logs are structured consistently and contain the context needed for analysis.

These capabilities complement the OpenTelemetry Collector by providing an easy way to manage enrichment and routing at scale. Organizations can continue using Datadog's extensive collection ecosystem, including integrations and agents, while delivering telemetry into ClickHouse using OpenTelemetry-compatible schemas.

The result is a flexible observability architecture optimized for high-volume telemetry. Teams can investigate recent data through Datadog while retaining complete telemetry datasets in ClickHouse, enabling historical analysis, long-range investigations, and access to full-fidelity data without changing workflows.  [ClickHouse Cloud’s separation of storage and compute](https://clickhouse.com/docs/cloud/reference/shared-merge-tree) allows organizations to retain that data at low cost, scale query resources independently, isolate workloads, and adapt capacity as requirements change without affecting retained data.

## Search ClickHouse logs directly from the Datadog Log Explorer

Storage is only useful if data remains accessible when it is needed.

With federated log search, teams can search logs stored in ClickHouse directly from the Datadog Log Explorer. Logs remain in ClickHouse, eliminating the need to duplicate or re-ingest data while preserving the workflows engineers already use for troubleshooting and investigation.

This approach provides several advantages.

First, organizations can retain significantly more telemetry for longer periods, ensuring historical data remains available when investigations require it. High-volume logs that might otherwise be sampled, filtered, or discarded can remain searchable through Datadog while being stored in ClickHouse.

Second, teams can work from a single interface and continue to get the experience they already know with Datadog, combined with the scale and full-fidelity telemetry retention of ClickHouse. Engineers no longer need to switch between multiple systems depending on where a particular dataset resides.

Finally, data stays where it was originally stored. Queries are executed against ClickHouse directly, avoiding the operational overhead of maintaining duplicate copies of telemetry.

## Bringing together the best of both platforms

By combining ClickHouse and Datadog, teams can store and retain far more telemetry data while continuing to investigate, troubleshoot, and search through the workflows they already know. The result is a flexible observability architecture: Datadog provides the collection, user experience, and investigation capabilities that engineering teams rely on every day, while ClickHouse provides a scalable foundation for long-term telemetry storage and analytics.

This partnership gives customers greater control over where their data lives, how long they retain it, and how they access it, without requiring them to compromise on performance, usability, or access to their data.

## Get started

The native ClickHouse integration for Datadog Observability Pipelines and federated log search from the Datadog Log Explorer are available in Preview.

If you're interested in combining ClickHouse's scalable observability storage with Datadog's observability experience, you can request access to the private preview today via [https://www.datadoghq.com/product-preview/federated-search/](https://www.datadoghq.com/product-preview/federated-search/).


Interested in seeing how ClickStack works on your observability data? Get started with Managed ClickStack in ClickHouse Cloud in minutes and receive $300 in free credits.

Get started today

    ClickHouse and Datadog are partnering to combine full-fidelity log retention at scale with the powerful search and investigation experience engineers rely on every day.

large_datadog-clickhouse-partnership.jpg

medium_datadog-clickhouse-partnership.jpg

small_datadog-clickhouse-partnership.jpg

thumbnail_datadog-clickhouse-partnership.jpg

datadog-clickhouse-partnership.jpg

Datadog and ClickHouse partner to bring full-fidelity data to modern observability

* QRT, a global investment manager, uses ClickHouse Cloud to power a data platform serving researchers, as well as a real-time risk and P&L system. 
* Beyond supporting researchers, ClickStack is also the foundation for QRT's observability infrastructure.

Summary

Quantitative trading is, at its core, a data challenge. Funds seek to store more of it, query it faster, and act on it first.

[Qube Research & Technologies (QRT)](https://www.qube-rt.com/) is a global quantitative investment manager headquartered in London. The firm's decision to use [ClickHouse Cloud](https://clickhouse.com/cloud) reflects a philosophy that dates back to the firm's origins. Most hedge funds founded in the 1990s and 2000s run on-premise data centers built up over decades. But QRT, launched in 2018 amid a rapid spinout of Credit Suisse's proprietary trading arm, was originally built in the cloud. Rather than inheriting legacy on-premise infrastructure, it built itself cloud-native from day one.

## Built for the cloud, ready to scale

QRT uses [ClickHouse Cloud](https://clickhouse.com/cloud) across two of their major systems. One centralized platform that gives researchers across the firm access to the data they need to develop trading strategies, and another near real-time risk monitoring, management, and P&L system.

"We wanted something faster and much more scalable," says a senior engineer on the team. "With our previous database, we were limited by the number of writers, the number of readers, and the total load per cluster. With ClickHouse, we don't have those limitations."

Beyond supporting researchers, ClickHouse is also the foundation for QRT's observability infrastructure. The migration to [ClickStack](https://clickhouse.com/clickstack), ClickHouse's integrated observability stack, consolidated logs, metrics, and traces onto a single backend for the first time. Alerts, log queries, and metric aggregations that used to require managing multiple systems now run against a single ClickHouse instance using standard SQL. "Having all our monitoring on one backend gives us the simplicity to maintain and is reliably fast," says the QRT engineer.

Interested in seeing how ClickHouse works on your data? Get started with ClickHouse Cloud in minutes and receive $300 in free credits.

How QRT, a cloud-native quantitative investment manager, uses ClickHouse Cloud and ClickStack to power firm-wide research, real-time risk and P&L, and observability at petabyte scale.

large_QRT Customer Story Cover.jpg

medium_QRT Customer Story Cover.jpg

small_QRT Customer Story Cover.jpg

thumbnail_QRT Customer Story Cover.jpg

QRT Customer Story Cover.jpg

How QRT powers real-time research and risk management at petabyte scale

_Veteran enterprise leaders join across APJ, public sector, financial services, solutions architecture, and revenue operations as the company scales its global organization_

**SAN FRANCISCO — June 8, 2026 —** ClickHouse, a leader in real-time analytics, data warehousing, observability, and AI/ML, today announced a significant expansion of its global go-to-market (GTM) leadership team, headlined by the appointment of **Ed Lenta** as **Vice President, Asia Pacific and Japan (APJ)**. The additions build on the momentum established last year with the appointment of **Kevin Egan** as **Chief Revenue Officer**, and reflect ClickHouse's strategic investment in scaling its global organization to meet surging customer demand.

Lenta joins ClickHouse to lead the company's go-to-market efforts across the APJ region. He brings deep experience scaling cloud and data platform businesses across the region, most recently as General Manager of Asia Pacific and Japan at Databricks, where he managed operations across more than twenty countries and led customer-facing teams serving organizations from startups to large enterprises and government. Earlier in his career, he helped build Amazon Web Services and VMware from early-stage companies into major platform providers in APJ. Based in Singapore, Lenta will be responsible for accelerating the adoption and expansion of ClickHouse Cloud and the open-source database in the region, empowering enterprises and public sector agencies to solve their growing data challenges.

"ClickHouse is emerging as the foundational data infrastructure for the AI era, and the strategic opportunity across Asia Pacific and Japan is enormous," said Lenta. "As agentic workloads scale, enterprises are rethinking their data architecture around both speed and efficiency, and ClickHouse outperforms traditional data warehouses by orders of magnitude on critical cost-performance metrics – that unique combination is what makes this technology so category-defining. I'm excited to build the teams and partnerships that will help organizations across the region unlock the full power of real-time analytics at scale."

Lenta's appointment in APJ is complemented by the recent hire of **Takeshi Kaneko** as **Country Manager of ClickHouse Japan**. Kaneko brings more than two decades of leadership in the Japanese technology market, including senior executive roles at Nutanix, where he served as President of Nutanix Japan, as well as leadership positions at Red Hat and Microsoft Japan.

The company also announced a series of additional GTM leadership appointments that strengthen its global revenue organization:

- **Billy Schoeffel, Vice President, Financial Services**, joins after more than six years leading global financial services sales at Snowflake, with prior senior sales leadership roles at ServiceNow and across the enterprise software industry.
- **Kenneth Melero, Vice President, Public Sector**, brings more than 30 years of experience across U.S. federal, state and local government, education, and worldwide public sector environments. He joins ClickHouse from Chainguard, where he led U.S. public sector sales, and previously served as a Regional Vice President at Elastic.
- **Maged Shehata, Vice President, Global Solutions Architecture**, brings more than 20 years of solution engineering experience in the data and AI space. Most recently at Snowflake, Shehata led solution engineering teams across retail and consumer goods, financial services, and healthcare, bringing expertise spanning data integration, data governance, real-time analytics, and broader enterprise data platforms.
- **Andrew Straus, Vice President, Global Revenue Strategy and Operations**, slated to join ClickHouse in late June, brings two decades of go-to-market and operating leadership across cloud infrastructure and enterprise software, spanning senior revenue operations and GTM strategy roles at Snowflake, Kong, and HPE.

"As organizations race to modernize their data infrastructure for the demands of agentic AI, we are investing decisively in the world-class team needed to serve them globally," said Kevin Egan, Chief Revenue Officer at ClickHouse. "Ed's leadership in APJ, combined with the strength and breadth of talent joining across the public sector, strategic sales, solutions architecture, and revenue operations functions, positions us to scale with our customers everywhere they operate."

The leadership expansion comes amid a period of exceptional momentum for ClickHouse. The company recently surpassed $250 million in annual run-rate revenue, more than triple a year earlier, and crossed 4,000 customers, adding more than 1,000 net new customers in a single quarter. The breadth of customers featured at the company's Open House 2026 user conference reflects how deeply ClickHouse now sits in the enterprise stack, with speakers including Visa, Cisco, Intuit, Shopify, DoorDash, Mercado Libre, Zoox, and Jump Trading. These leadership appointments are a direct investment in that growth, adding the regional coverage, enterprise relationships, and operational infrastructure to support customers at scale.

<br />

<div align="center">
  <img src="https://cms-new.clickhouse.com/uploads/ed_lenta_headshot_7c4a3d6abe.jpg" alt="Ed Lenta" width="400" />
</div>

ClickHouse, a leader in real-time analytics, data warehousing, observability, and AI/ML, today announced an expansion of its global go-to-market (GTM) leadership team, including the appointment of Ed Lenta as Vice President, Asia Pacific and Japan (APJ).

large_clickhouse-appoints-apac-leader.jpg

medium_clickhouse-appoints-apac-leader.jpg

small_clickhouse-appoints-apac-leader.jpg

thumbnail_clickhouse-appoints-apac-leader.jpg

clickhouse-appoints-apac-leader.jpg

ClickHouse appoints new leader for Asia Pacific and expands global go-to-market leadership team

If you missed Day 1, catch up on all the ClickHouse Cloud and ClickStack announcements [in our Day 1 roundup blog post](https://clickhouse.com/blog/open-house-2026-day-1).

Last year was our first annual Open House conference, and we covered a lot in a single day. This year, we had so much content and announcements to make that we decided to extend by another day. Day 2 delved deeper into what the open-source community, our partners, and our customers are building on top of ClickHouse. We focused not just on the features being built, but also on why the world is betting on this data platform.

This year, ClickHouse turns 10 as an open source project. Over that decade, the ecosystem has grown to 200+ integrations built by us, by partners, and by a community that just keeps shipping. The pace has accelerated recently. MCP support and our skills framework are seeing explosive adoption, and we are watching entirely new categories of embedded and bespoke integrations emerge organically. People and agents alike are finding ways to wire ClickHouse into their stacks that we would not have thought to build ourselves.

## 10 years of Open Source {#10-years-of-open-source}

The opening keynote looked back at 10 years of ClickHouse as an open-source project, including contributor growth, community milestones, and the launch of the [Community Champions Program](https://clickhouse.com/community). Alexey Milovidov took the stage and walked us through his thought process in building and designing ClickHouse and what he has learned over the years in a session entitled 'How to build a great database.' It hasn't just been Alexey working on ClickHouse all this time. With over 2.6k contributors to ClickHouse and nearly 48k stars in GitHub, the support we've seen from our community over the last decade has been amazing, and we can't thank you enough for all of your contributions.

![open_house_day2_may2026_image5.png](/uploads/open_house_day2_may2026_image5_54d9436e8d.png)

## House Mates: the inaugural cohort {#house-mates-the-inaugural-cohort}

To support the momentum behind the ClickHouse ecosystem, we announced [House Mates](https://clickhouse.com/blog/introducing-house-mates), ClickHouse's official partner program. House Mates was officially launched with 25+ ISV & technology partners and 35+ services & channel partners. You can't ask for more out of our first cohort; they shipped.

The first wave of House Mates covers the integrations our customers have asked about most. On the data and transformation side, Fivetran is now generally available as a ClickHouse Cloud destination, bringing 500+ sources with schema migration support and an enterprise SLA. Sigma Computing is moving its official connector to public beta. Notion is adding first-party support for the ClickHouse MCP server for custom agents. And ClickHouse is now the first partner to build a community dbt Fusion adapter, taking our first step towards natively running on the dbt platform. More on each of these below.

Beyond those announcements, the cohort spans: Artie for sub-minute CDC replication from Postgres, MySQL, SQL Server, Oracle, and MongoDB; Dreambase as an AI-native analytics platform connecting ClickHouse and Supabase; Laravel for a full-featured ClickHouse driver with Eloquent models, Query Builder, Schema Builder, and native migration support via artisan; Odigos for zero-code OpenTelemetry tracing with ClickHouse as the analytics backend; Tavily for AI search with ClickHouse Agents as a fast retrieval backend; Hightouch for reverse ETL from ClickHouse to CRMs, ad platforms, and SaaS destinations; and DataHub as an open source data catalog with lineage and discovery for ClickHouse tables and pipelines. The list goes on. 

We're humbled to have such an amazing group of partners building for tomorrow with us. During our conference, we had some exciting announcements across the ClickHouse ecosystem.

## Fivetran: generally available {#fivetran-generally-available}

The Fivetran destination connector for ClickHouse Cloud is now generally available. This is one of the most-requested integrations we have heard from customers over the past two years.

Teams can now replicate data from 500+ sources into ClickHouse without writing or maintaining pipeline code. Fivetran handles schema migrations, column additions, and retries automatically. The GA release includes schema migration support and an enterprise SLA, making this a production-ready path for teams running enterprise SaaS data into ClickHouse from Salesforce, HubSpot, Google Ads, Stripe, SAP, NetSuite, Snowflake, Databricks, and hundreds more.

> "ClickHouse has been a fast-growing data destination on Fivetran, and going GA reflects how seriously both teams have invested in this partnership. Enterprise data teams run their businesses on sources such as Salesforce, Google Analytics, SAP, Databricks, Snowflake, Workday, and NetSuite. Getting that data into ClickHouse reliably, with automated schema management and an enterprise SLA, is what Fivetran was built for. We are proud to give customers a production-grade path from every major enterprise source into ClickHouse Cloud."
> 
> -- Shiva Mogili, Director of Product Management, Connectors & Extensibility

**Google Cloud Dataflow: Pub/Sub template generally available**

Adding to our BigQuery template, the ClickHouse template for Pub/Sub support is now in the Google Cloud Dataflow gallery. 

Previously, moving Pub/Sub data into ClickHouse required custom Beam pipelines with no official support. Now teams can configure a managed pipeline to ClickHouse directly from the Dataflow console. For teams on ClickHouse Cloud, [the same integration is also available as a native ClickPipes connector](https://clickhouse.com/blog/open-house-2026-day-1#clickpipes), so you can set it up from whichever side makes more sense for your workflow.

Learn more about how to get started using Pub/Sub on Dataflow [here](https://clickhouse.com/docs/integrations/google-dataflow/templates/pubsub-to-clickhouse).

## Sigma Computing: connector in public beta {#sigma-computing-connector-in-public-beta}

The Sigma Computing connector for ClickHouse is moving from private beta to public beta after successfully being used in production across a variety of customers. The connector is now available to any Sigma customer who wants to connect to ClickHouse.

In addition to the connector going to public beta, Sigma also selected ClickHouse as the first database target for its new ETL Cache Layer, now in private preview. When Sigma queries hit the primary warehouse at every dashboard load, teams pay full compute cost regardless of how often the data underneath is actually changing. The ETL Cache Layer pre-materializes hot datasets and serves them sub-second at any concurrency level, with intelligent routing that falls back to the source warehouse on a cache miss. This has not only major performance benefits, but cost benefits as well.

*"ClickHouse is one of the fastest-growing databases in the market, and our customers are building on it because it delivers disruptive cost-performance," said Mike Palmer, CEO, Sigma. "Entering the House Mates partner program is our commitment to making Sigma the best AI runtime experience on top of ClickHouse."*

## dbt: ClickHouse Fusion adapter {#dbt-clickhouse-fusion-adapter}

ClickHouse is the first partner to co-develop a dbt Fusion engine adapter with dbt Labs. Fusion is a true SQL compiler built in Rust: the engine underneath dbt that makes it faster at any project scale, gives developers real feedback before anything hits the warehouse, and generates the structured metadata that makes AI-assisted data work reliably. It's the future of dbt, and we are excited to partner with dbt Labs to bring the adapter to our joint community.

Our growing community of dbt users, combined with the dbt vision, was the catalyst for this work. As a result, the adapter is now available in alpha via the CLI. Analytics engineers can run dbt models against ClickHouse using Fusion today. This is the first milestone toward a full dbt platform (fka Cloud) integration. We're excited not only to address one of our most common integration requests but also to help shape the new adapter program at dbt Labs. As Fusion becomes the default, we're committed to delivering a first-class experience for our shared customers.

> "ClickHouse has become one of the fastest-growing databases in the dbt community, and it is easy to see why. The combination of speed at scale and a genuine commitment to the open ecosystem makes it a natural fit for the composable data infrastructure the market is moving toward. Fusion is how dbt will power the platform going forward, and getting ClickHouse there early means our shared customers can build a fully open stack today. We are excited about what we are building together, and there is a lot more to come."
> 
> -- Hope Watson, Product Manager, dbt Labs

## Apache Airflow: native provider {#apache-airflow-native-provider}

A native Apache Airflow provider for ClickHouse will be available from the Airflow registry [soon](https://github.com/apache/airflow/pull/67080). We've seen a lot of interest in native orchestration connectors and want to thank Ivan Klimenko ([klimenkoIv](https://github.com/klimenkoIv)), Anton Bryzgalov ([bryzgaloff](https://github.com/bryzgaloff)), and the Airflow community for [supporting this effort](https://github.com/bryzgaloff/airflow-clickhouse-plugin) to date. Our native provider adds a ClickHouse hook and operator to the Airflow ecosystem, so teams can schedule queries, table refreshes, and data loads directly from DAGs without writing custom operator wrappers. The new provider is built on the new Apache Airflow standards, will be maintained by ClickHouse, and will work out of the box with Astronomer Astro connections once it clears the registry. We're excited to see how this project evolves and can't wait to get this out to our community.

## Vercel AI SDK v7 + Langfuse {#vercel-ai-sdk-v7-langfuse}

Vercel AI SDK v7 ships with a new Telemetry system for more granular agent observability. Langfuse offers a native integration with this new telemetry system via OpenTelemetry. 

AI SDK v7 is available as a canary build today and will be generally available in June. 

Langfuse integrates with the AI SDK v7 with a single `registerTelemetry()` call, giving full hierarchical trace visibility across every tool call, sub-agent span, and LLM invocation in a production AI agent. No extra boilerplate, no custom instrumentation wiring.

![open_house_day2_may2026_image4.png](/uploads/open_house_day2_may2026_image4_7d04c525ed.png)

## Notion: Native ClickHouse connector for custom agents {#notion-native-clickhouse-connector-for-custom-agents}

Notion is adding ClickHouse as a native connector for custom agents, so teams can query their ClickHouse data directly from inside Notion without any external tooling. 

What makes this integration interesting is the architecture. There's no separate semantic layer sitting between your data and your questions. Instead, Notion pages written in plain English describe your table structures, how data is tagged, and what different fields mean. The agent reads those docs, connects to ClickHouse via MCP, and returns answers in natural language. Your existing Notion workspace becomes the context layer. ClickHouse does the querying. This will be rolling out to Notion custom agents soon.

> "Every day, ClickHouse users rely on a service that handles some of the most demanding data workloads out there. But getting those insights into the tools where teams actually make decisions has historically required a lot of glue. We're thrilled that Notion custom agents can now query ClickHouse directly: our shared customers can ask questions in plain language and get real, data-rich answers back, no SQL required. That means scheduled reports, automated workflows, and live data pulled straight into docs and dashboards. ClickHouse crunches the numbers, Notion makes them actionable."
> 
> -- David Rosenberg, Ecosystem Lead, Notion

## Python Client is now v1! {#python-client-is-now-v1}

[clickhouse-connect v1](https://github.com/ClickHouse/clickhouse-connect) is now generally available on PyPI. This is a huge milestone and the first major release of the client. The headline for the initial 1.0 release is a native async client built from the ground up on top of aiohttp. This replaces the old executor-based wrapper and has full feature parity with the sync client, including streaming.

You'll also find massive speedups around performance for DateTime types, fixed-width numerics, Maps, Decimals, and BigDecimals. Along with support for Variant columns that can now be serialized client-side using their native ClickHouse types.

For projects that have numpy, pandas, pyarrow, or polars installed, cold-start time improves by 4x. That improvement comes from lazy-loading optional dependencies rather than importing everything up front. 

We have introduced first-class support for the SQLAlchemy dialect, deepened support for ClickHouse-specific SQL, and paved the way for Alembic schema migrations, which ship in v1.1.0.

On the compatibility front, this release also adds compatibility for Pandas 3.x in addition to experimental support for Python 3.14free-threading, giving existing clickhouse-connect users compelling reasons to upgrade

## .NET Stack: generally available on NuGet {#net-stack-generally-available-on-nuget}

With the recent stable release 1.2.0, [clickhouse-cs](https://github.com/ClickHouse/clickhouse-cs) now ships a full .NET development stack with four packages on NuGet.

[ClickHouse.EntityFrameworkCore](https://github.com/ClickHouse/ClickHouse.EntityFrameworkCore) adds ORM support, including JOINs, UNIONs, subqueries, DDL generation, and schema migrations. [Serilog.Sinks.ClickHouse](https://github.com/ClickHouse/Serilog.Sinks.ClickHouse) ships with column and cluster fluent configuration for structured logging into ClickHouse. [ClickHouse.Aspire](https://github.com/ClickHouse/ClickHouse.Aspire/) rounds out the stack for teams building cloud-native .NET applications.[ClickHouse.SemanticKernel](https://github.com/ClickHouse/ClickHouse.SemanticKernel/) connects ClickHouse to the Microsoft AI orchestration ecosystem. Each package is available today for users building in the .NET ecosystem.

## ClickStack announcements {#clickstack-announcements}

Following up on yesterday's ClickStack Cloud announcement, we also launched two major additions to ClickStack: **AI Notebooks** and the **ClickStack MCP server**. Both are built around a pretty simple idea: observability workflows shouldn't feel isolated from the way engineers actually debug systems today.

**AI Notebooks**, now public beta in Managed ClickStack, give teams a persistent workspace for investigations. Instead of losing context across dashboards, chats, terminals, and ad hoc queries, engineers can keep queries, charts, notes, and findings together in one place while an investigation evolves.

Most production incidents aren't linear. You follow one lead, hit a dead end, revisit an earlier assumption, compare traces against logs, then try something else. Notebooks are designed to support that kind of workflow without forcing everything into a single chat session or rigid sequence of steps.

![open_house_day2_may2026_image1.png](uploads/open_house_day2_may2026_image1_a745b3bbd1.png)

At the same time, we're seeing more teams build their own internal agents and automation around observability data using tools like Claude Code, Cursor, and custom SDK-based systems. The **ClickStack MCP** server is designed for that world and built around the philosophy of "bring your own agents".

Rather than introducing yet another standalone AI interface, the MCP server exposes the same observability tools used internally by Notebooks, but makes them available to external agents. Under the hood, the MCP translates those operations into optimized ClickHouse queries, allowing agents to work with higher-level semantic endpoints instead of repeatedly rebuilding investigation logic and generating complex SQL from scratch.

![open_house_day2_may2026_image3.png](/uploads/open_house_day2_may2026_image3_d527806633.png)

Compared to the more generic ClickHouse MCP implementation, our internal testing shows the ClickStack MCP server improves efficiency, performance, and investigation accuracy. In our internal testing, investigations were completed up to 25% fewer tool calls, a 2.5x increase in consistency, and an improvement in evaluation of almost 20%.

For more details, including where we're heading with tighter integration between Notebooks and MCP-driven workflows over time, check out the [main announcement post](https://clickhouse.com/blog/observability-mcp-server-ai-notebooks).

## What this adds up to {#what-this-adds-up-to}

Ten years in, the ecosystem around ClickHouse is healthier than it has ever been and expanding rapidly. The integrations we announced today span ELT, BI, orchestration, language clients, MCP, and AI observability. They close gaps that have been on our list for a long time and open up entirely new categories of things people (and their agents) can build.

The community, our partners, and our customers are all pulling in the same direction. That is what Day 2 was about, and we can't wait to see what you build.

Session recordings will be available shortly. For everything announced on Day 1, start here: [Day 1 ClickHouse Cloud recap](https://clickhouse.com/blog/open-house-2026-day-1).


A Day 2 recap of Open House 2026 covering ClickHouse's 10 years of open source, the House Mates partner program launch, and ecosystem announcements from Fivetran, Sigma, dbt, Notion, Replit, and Google Cloud.

large_open_house_day2_may2026_image2.png

medium_open_house_day2_may2026_image2.png

small_open_house_day2_may2026_image2.png

thumbnail_open_house_day2_may2026_image2.png

open_house_day2_may2026_image2.png

Open House 2026 Day 2: 10 years of open source and what the ecosystem is building next

ClickHouseがお客様のデータでどのように動作するか興味がありますか？ClickHouse Cloudを数分で開始でき、$300分の無料クレジットを受け取れます。

今すぐ始める

Open House 2026 の 2 日目を振り返り、ClickHouse のオープンソース 10 周年、House Mates パートナープログラムの開始、そして Fivetran、Sigma、dbt、Notion、Replit、Google Cloud によるエコシステム関連の発表を取り上げます。

Open House 2026 2日目: オープンソース10年の歩みと、エコシステムが次に築くもの

- ClickHouse Open House 2026 Day 1 brought announcements across ClickHouse Cloud, Postgres, distributed query execution, AI agents, observability, and lakehouse interoperability.
- ClickHouse Postgres moves to public beta delivering over 5x more transactions per second than AWS RDS; multi-stage distributed queries cut TPC-H SF100 runtime from 117.6 seconds to 54.7 seconds; Langfuse V4 delivers 200x query performance improvements.
- ClickHouse Agents powered by Claude entered public beta with a native chat experience and no-code agent builder; enabling bi-directional workflows between ClickHouse and external lakehouse systems.

Today at Open House 2026 in San Francisco, we shared a broad set of announcements across ClickHouse Cloud, AI agents, observability, Postgres, and lakehouse interoperability.

Open House has always been a chance for our users to come together and exchange ideas about where real-time data systems are heading, and this year’s launches reflect many of the conversations we’ve been having with users, contributors, customers, and partners over the last year. This post rounds up the major announcements from Day 1, and tomorrow we’ll be back with another set of announcements from Day 2.

## Improving the UX for humans and agents {#improving-the-ux-for-humans-and-agents}

ClickHouse Cloud is now serving two audiences: the teams that operate it in production and the AI agents that query it programmatically. These product updates bring improvements across resilience, observability, schema management, and developer tooling to make ClickHouse Cloud easier to operate and easier to build on.

Cross-Region Replication brings an active-passive failover architecture to ClickHouse Cloud. Data is synchronously replicated to a secondary region, with recovery time measured in minutes and recovery point measured in seconds. We will begin a limited private preview later this year.

Monitoring v2 overhauls observability with opinionated service health signals, schema exploration, query insights, and mutation tracking all in one place.

A Materialized View Pipeline Visualization will give teams a live visual map of data flowing across materialized views, one-click debugging for unhealthy refreshes, and a drag-and-drop pipeline builder for creating new pipelines.

![](/uploads/open_house_day1_may2026_image3_ed975af860.png) 

*Coming soon* Schema Management and Optimization introduces a fully guided, end-to-end AI-assisted schema management experience. It includes a workload-aware recommendation engine, a dedicated sandboxing environment, automated impact analysis, and a guided blue-green deployment flow for zero-downtime schema changes.

![](/uploads/open_house_day1_may2026_image10_182df22dd5.png)

On the developer and agent side, the ClickHouse CLI is built for agentic workflows, with official ClickHouse skills installable directly into Cursor, Claude Code, and other coding agents. The same CLI manages both local development and production Cloud services.

Query API Endpoints gets a unified management pane, enterprise-grade RBAC, native IDE integrations, and MCP tooling support. The AI-Enhanced Query Builder UI lets analysts build complex queries visually and move fluidly between the visual builder and ClickHouse Assistant.

And MCP-as-a-Service enables fully managed, domain-specific remote MCP servers to be spun up directly from the ClickHouse Cloud UI, with fine-grained access control and support for custom context.

Together, these updates make ClickHouse Cloud a more complete platform for running production workloads, whether the queries come from a dashboard, a notebook, or an agent.

## Postgres announcements {#postgres-announcements}

AI workloads are collapsing the divide between transactional and analytical databases. Applications that once ran predictable, hard-coded queries now generate unpredictable bursts of agent-driven requests that need answers from both sides of the stack. Best-of-breed matters more than ever: Postgres for OLTP, ClickHouse for OLAP. But composing them has traditionally meant standing up each database independently and bridging the two with custom CDC pipelines, message brokers, and orchestration logic.

ClickHouse Postgres, now moving from private preview to public beta, is built to close that gap. The service runs on local NVMe storage, eliminating the primary bottleneck in transactional workloads by colocating storage with compute instead of relying on network-attached volumes.

In early benchmarks, this delivers over 5x more transactions per second than AWS RDS and 2.4x more than the next closest alternative. High availability is supported with up to two standby replicas across availability zones, synchronous commit to the fastest standby, and automatic failover. Continuous WAL archiving to S3 provides point-in-time recovery, branching, and region-survivable durability.

![](/uploads/open_house_day1_may2026_image8_904253765c.png)

On the integration side, a Postgres-native CDC pipeline streams inserts, updates, and deletes directly into ClickHouse with no intermediate infrastructure, handling both parallel initial snapshots and continuous replication.

To make the experience even simpler for developers already building on Postgres, a new open-source extension, pg_clickhouse, makes ClickHouse-backed tables queryable directly from within a standard Postgres session, transparently pushing projections, filters, and aggregations down to ClickHouse for fast analytical queries. Developers continue using familiar Postgres SQL and tooling while the right engine handles the right workload behind the scenes.

For a deeper look at the architecture, benchmarks, and roadmap, read the [dedicated blog post](https://clickhouse.com/blog/postgres-managed-by-clickhouse-beta).

## Distributed queries in ClickHouse {#distributed-queries-in-clickhouse}

ClickHouse Cloud now supports multi-stage distributed query execution in private preview. The feature scales large joins and high-cardinality aggregations across many nodes by repartitioning intermediate data between execution stages.

Parallel replicas already distribute probe-side work, but every node still rebuilds the full hash table from the right side of the join. High-cardinality aggregations hit a similar wall, with final aggregation collapsing onto a single coordinator. Multi-stage execution removes both bottlenecks.

![](/uploads/open_house_day1_may2026_image1_73ff57b040.png)

Workers exchange intermediate results, re-partition by join or GROUP BY keys, and operate on independent slices of the problem.

The feature builds on ClickHouse Cloud's shared-storage architecture, where any worker can access any data without fixed shard ownership. That makes it possible to distribute query stages dynamically across available compute, pointing toward more elastic execution with interchangeable worker pools.

In early TPC-H testing at scale factor 100, the full benchmark suite ran in 54.7 seconds on eight nodes, down from 117.6 seconds on a single node. Roughly a 2x improvement, with further gains expected as the cost-based optimizer matures.

For more technical details, we recommend users read the [dedicated blog post](https://clickhouse.com/blog/multi-stage-distributed-query-execution-clickhouse-cloud).

## **Join performance and other core improvements**

We presented how JOIN performance in ClickHouse improved by over 6x in the last year through support for correlated subqueries, lazy materialization, runtime filters, and automatic join reordering. These optimizations and features now put ClickHouse on par with established data warehouses across TPC-H and other standard JOIN benchmarks, often matching or exceeding the performance of systems such as Snowflake, BigQuery, and Databricks.

![image5.png](/uploads/image5_760140b60c.png)

We also recapped some of the most recent improvements in the core database, including Full-Text Search being made [generally available](https://clickhouse.com/blog/full-text-search-ga-release), addressing one of the most common requirements for observability and AI workloads. 

## ClickPipes {#clickpipes}

We're bringing ClickPipes natively to GCP! Starting today, new ClickHouse Cloud services on GCP run ClickPipes in the same region for lower-latency ingestion, data locality guarantees, and tighter integration with GCP-native features.

This includes Private Service Connect (PSC) support for secure, direct private connections to GCP-managed services behind a VPC; before, you had to tunnel connections via an SSH bastion host.

Note: If you’re using ClickPipes with existing GCP services, please reach out to your account executive to discuss a migration plan.

![](/uploads/open_house_day1_may2026_image9_9179adfb9e.gif)

To deepen our integration with the GCP ecosystem even further, we’re also announcing a new ClickPipes connector for Google Cloud Pub/Sub, in Private Preview. If Pub/Sub is part of your stack, you can now subscribe to topics and stream data directly into ClickHouse Cloud with no additional infrastructure. The connector supports all common formats (JSON, Avro, Protobuf) and schema registry integration, with attribute-based message filtering, flexible seek options, and per-key ordered delivery. Like all connectors, it can also be managed programmatically via OpenAPI and the ClickHouse Terraform provider.

## Bring any agent to your real-time data, with zero lock-in {#bring-any-agent-to-your-real-time-data-with-zero-lock-in}

AI agents are a new kind of workload, and they're breaking the old playbook. They fire ten to a hundred times the volume of queries a human analyst would, in bursts, all at once, and most analytical engines simply weren't built for it. Closed data platforms make it worse, locking teams into limited and proprietary models, a single agent, and one rigid way of working, while costs spiral. The agentic era needs a real-time substrate that's fast, affordable, and refuses to lock anyone in.

![](/uploads/open_house_day1_may2026_image2_2a92f8d9fc.png)

Today, we're introducing ClickHouse Agents, a fully managed agentic analytics service on ClickHouse Cloud, powered by Claude and now in public beta.

Every ClickHouse Cloud user gets a native chat experience and a no-code agent builder for shipping agents grounded in their own data, with zero setup. Under the hood is a sandboxed code interpreter, shareable artifacts, skills, memory, and multi-agent workflows, all built on LibreChat - the open-source platform that [joined ClickHouse last November](https://clickhouse.com/blog/clickhouse-acquires-librechat).

Because it speaks MCP natively, context can be pulled from any MCP-compatible system. ClickHouse is purpose-fit for this agentic era, where sub-second queries on billions of rows, petabyte scale, and cost efficiency are key to running agentic analytics in production successfully. Customers don't have to choose between us and the tools they already love. Use our agents out of the box, build or bring your own, get speed, transparency, and full control with your analytics AI stack.

## Langfuse announcements {#langfuse-announcements}

As we discussed in the previous section, AI agents are introducing a fundamentally different workload pattern for data systems, and observability platforms are feeling that shift just as strongly. A single agent run can produce hundreds of observations across LLM calls and tool calls, and the interesting signal is rarely at the top of the trace but requires deep evaluation workflows in production and in development. Langfuse solves for this problem, and the team [joined ClickHouse in January](https://clickhouse.com/blog/clickhouse-acquires-langfuse-open-source-llm-observability).

Today, we're announcing Langfuse V4, in beta on Langfuse Cloud and soon to be available for self-hosted deployments.

This represents a "simplify for scalability" rewrite that collapses the data model into a single immutable observations table with no joins and no deduplication. The result is millisecond dashboard loads, 200x query performance across many important routes, and a single project that comfortably handles billions of events, all benefiting from ClickHouse's performance. Learn more about the details of this change in this [technical blog post](https://langfuse.com/blog/2026-03-10-simplify-langfuse-for-scale).

On top of V4, we're shipping a tighter improvement loop: Experiments now ship with an improved UI, code-based evaluators, and CI/CD workflows. This means prompt and agent changes can be scored before they reach production, with LLM-as-a-judge supporting categorical, boolean, and free-text scores and alerting, ensuring teams can react to score and latency regressions at agent scale.

The platform itself is becoming agent-native, with v2 of the MCP server exposing most API routes to agents, a rewritten Langfuse CLI, and skills that wrap every Langfuse capability so the same building blocks work for humans and agents. Soon, these APIs will include fast full-text search, allowing agents to explore traces in a more free-form/semantic way.

Finally, Langfuse Cloud users can now sign in using their ClickHouse Cloud identity, making adoption seamless and removing the need to manage a separate login. Langfuse itself remains open source, OpenTelemetry-native, self-hostable, and built to scale with the realities of agent workloads.

## ClickStack announcements {#clickstack-announcements}

### ClickStack Cloud {#clickstack-cloud}

Today, we announced the private preview of ClickStack Cloud, a fully managed, serverless observability platform built on ClickHouse. Teams can send OpenTelemetry data to a managed OTLP endpoint, then investigate logs, metrics, and traces through the ClickStack UI without operating collectors, ingestion pipelines, or ClickHouse clusters themselves. ClickStack Cloud enters private preview with managed ingestion, a serverless query experience, and integrated observability workflows built directly on ClickHouse Cloud.

During the private preview, we’re focused on two areas that matter for large-scale observability workloads: independently scaling ingestion and query infrastructure, and automatically tuning the underlying datastore based on how teams actually use their telemetry data.

This means improving systems that learn from common query patterns and automatically optimize telemetry data over time. Planned areas of automatic tuning include materializing frequently queried fields, adjusting primary keys around common filters, and adding materialized views and indexes for common dashboard and investigation workflows. These capabilities will be refined alongside early adopters and design partners throughout the private preview period.

Private preview spots are limited. If you are interested in using ClickStack Cloud, sign up for the [preview program](https://clickhouse.com/cloud/clickstack-cloud-waitlist) and tell us about your observability workload.

![](/uploads/open_house_day1_may2026_image6_0cf7b01fd7.png)

For more details, see our [dedicated blog post](https://clickhouse.com/blog/clickstack-cloud-private-preview).

### Managed ClickStack {#managed-clickstack}

In addition to ClickStack Cloud entering private preview, our existing Managed ClickStack offering is now generally available.

Managed ClickStack is designed for teams that want direct operational control over their observability stack, including ingestion pipelines, compute sizing, workload isolation, schema design, and datastore tuning. Users manage their own OpenTelemetry collectors and ingestion architecture while using ClickHouse Cloud as the underlying observability datastore. For many large-scale deployments, that control is essential for optimizing performance and achieving market-leading cost efficiency.

![image8.png](/uploads/image8_75f25bb9a6.png)

### PromQL support

While we believe SQL is the lingua franca of data analysis, some workloads benefit from a domain-specific language. PromQL for Prometheus-style metrics is one example.

To close out our observability announcements, we wanted to share an early preview of a major area of investment for 2026: PromQL support in ClickHouse and ClickStack.

During the demo, we showed PromQL queries running against ClickHouse through the ClickStack UI.

![promql.png](/uploads/promql_9967e55514.png)

This is still very early work. There are dragons here. The implementation is incomplete, behavior will change, and there is still a large gap between the current state and something we would consider production-ready.

For anyone curious enough to experiment, the underlying functionality is already available experimentally in ClickHouse, with integration also available in open-source ClickStack.

Most of the recent work has focused on language coverage and compatibility. The goal is straightforward: existing PromQL queries should behave the way users expect when pointed at ClickHouse. Performance work is happening alongside this, so PromQL queries can still benefit from the scale and execution speed of the ClickHouse engine.

## Partner program announcement {#partner-program-announcement}

ClickHouse also introduced House Mates, its first formal partner community and program. It launches with a founding cohort of more than 25 technology partners and over 35 services, consulting, and channel partners across six continents. The program is organized across three tracks: Technology, Services, and Reseller, each with three tiers: Ignite, Accelerate, and Prime. Benefits scale with tier and include joint go-to-market motions, co-innovation and integration support, enablement and certifications, incentives, and a dedicated partner portal. [Read the announcement blog](https://clickhouse.com/blog/introducing-house-mates) for more details. 

![image6.png](/uploads/image6_39ae0d5084.png)

## Wrap-up {#wrap-up}

Today’s announcements span distributed query execution, observability, lakehouse interoperability, agentic analytics, and transactional workloads, all while keeping teams in control of the tools and architectures they use. Whether that means bringing your own agent, your own catalog, your own collectors, or continuing to work from familiar Postgres workflows, the focus remains the same: fast real-time systems without lock-in. 

The performance improvements behind these launches were equally significant, including over 5x higher throughput than AWS RDS for Postgres workloads, roughly 2x faster distributed query execution at TPC-H scale factor 100, and up to 200x query performance improvements in Langfuse V4. We’ll be back tomorrow with another round of announcements and deeper technical sessions for Day 2.

A full recap of Day 1 at Open House 2026, covering every major announcement across ClickHouse Cloud, Postgres, distributed queries, ClickPipes, ClickHouse Agents, Langfuse V4, ClickStack, and the House Mates partner program.

ClickHouse

Blog posts by ClickHouse

Datadog and ClickHouse partner to bring full-fidelity data to modern observability

How QRT powers real-time research and risk management at petabyte scale

ClickHouse appoints new leader for Asia Pacific and expands global go-to-market leadership team

Open House 2026 Day 2: 10 years of open source and what the ecosystem is building next

Open House 2026 2日目: オープンソース10年の歩みと、エコシステムが次に築くもの

Open House 2026 Day 1: Real-time data without lock-in and what teams can build next