ClickHouse is shifting from a fast OLAP database to an AI data layer. The homepage now leads with “The leading database for AI,” and the top navigation ties together ClickHouse Cloud, BYOC, Postgres managed by ClickHouse, Managed ClickStack, Agentic Data Stack, and Langfuse Cloud. That is not a narrow database pitch, it is a platform bet. Homepage.

The recent product posts match that positioning. ClickHouse just published pg_clickhouse updates covering JSONB support, SQL value-function pushdown, array function pushdown, and HTTP result-set streaming. It also announced Terraform and OpenAPI support for ClickPipes as generally available, with full connector coverage and CDC support for Postgres, MySQL, and MongoDB.

The search story is even more aggressive. In its Elasticsearch comparison post, ClickHouse says a redesigned full-text search layer built on inverted indexes closes the old split between search and analytics for log workloads. The company claims 2 to 6x faster runtime on a realistic OpenTelemetry workload across datasets up to 50 billion rows. That is a vendor benchmark, not a universal law, so treat it as directionally useful, not final truth.

The shipping signal is strong. ToolVitals shows 62 release events in 30 days, 30 GitHub releases in 90 days, 47,073 GitHub stars, and a perfect 100 health and shipping score. The GitHub repo shows 227,545 commits, 4.8k issues, and 1.1k pull requests, plus a tagged v25.12.10.7-stable release from Apr 24. Compared with Weaviate’s 12 release events in 30 days and OpenViking’s 10, ClickHouse is moving much harder than the other database names in the tracker.

What the data does not tell you is whether the query planner is clean, the docs are good, or customers are happy. ToolVitals sees releases, stars, uptime, and SSL. It does not see code quality, revenue, retention, support burden, or how painful the system is to operate in production.

If your team is building agent-heavy analytics, observability, or PostgreSQL-plus-warehouse workflows, ClickHouse deserves a fresh look. The company is betting that storage, ingestion, search, and AI observability should sit close together, and the release velocity says they are serious about that bet.

Sources