ClickHouse has opened a public beta for a managed observability service that accepts OpenTelemetry data and handles ingestion, schemas and storage. It also has a private preview of AI-built observability apps, letting teams describe custom ways to investigate their live data.
ClickHouse Watch analysis
What happened
The public beta is powered by ClickHouse Cloud and open-source ClickStack. Teams send telemetry to a managed endpoint, then explore logs, metrics and traces in the ClickStack interface. Agents can also investigate through purpose-built tools and SQL. ClickHouse says teams do not need to size compute or tune the database and schemas themselves.
During the beta, logs and traces cost $0.10 per GB ingested, while metrics cost $0.05 per million data points. Those are half the listed standard prices of $0.20 per GB and $0.10 per million data points. Beta pricing includes 30 days of retention; longer retention is planned for general availability.
The separate Observability Apps preview lets users describe an application in natural language. An agent generates queries, visualisations and connected screens over live data, which users can refine and share. ClickHouse says generated code runs in a sandbox and queries are read-only, using the viewing user’s database permissions.
Why it matters
Observability tools can make teams adapt their questions to a vendor’s dashboards. ClickHouse is betting that managed telemetry plus custom, agent-built interfaces can reverse that bargain, without every team needing to build its own observability product. The beta’s volume-based pricing also gives teams a clearer starting point than estimating compute and storage needs, though actual costs still depend on how much data they send.
Our read
This is a practical combination: fewer database chores for teams that want managed service, and more room to tailor investigations than a fixed set of dashboards usually allows. The distinction between the public beta and the private-preview apps matters. The former is available to try; the more novel AI-generated interface is not yet generally available. And a generated investigation is only as useful as the data and queries behind it. Agents have been invited into the control room, but the humans still need to check the instruments.
What to watch
- How automatic scaling and schema optimisation improve during the beta.
- Whether ClickHouse publishes clearer guidance on costs as usage grows.
- When Observability Apps leave private preview, and how well generated apps handle real investigations.
- Whether teams prefer managed ClickHouse Observability or self-configured ClickStack, which remains available and open source.
Discussion spark: Would you trust an agent to build the observability interface your team uses during an incident, or should that interface stay fixed and carefully controlled?
Sources and evidence
- Source update (9 November 2026, 16:56 UTC)
Independent WittyWires tracker for public updates about ClickHouse. Not affiliated with or endorsed by ClickHouse; this is not an official account.