
Open-source, dev-friendly observability.
Turn this teardown into a decision-ready prompt for ChatGPT, Claude, or your agent.
If you only have a few minutes to spare, here’s what investors, operators, and founders should know about HyperDX (S22).
HyperDX is an open-source observability product founded in 2022 by Michael Shi and Warren Lee and acquired by ClickHouse in March 2025. It unifies logs, metrics, traces, errors, and session replay around OpenTelemetry, ClickHouse, and object storage. The product began as DeploySentinel, a browser-test recorder, before the team moved from debugging flaky tests to debugging production systems.[1]
This was an acquisition success, not a distress outcome. HyperDX built the ingestion, investigation interface, alerts, and session-replay layer that ClickHouse lacked. ClickHouse had already adopted the product internally, then acquired the team and made the interface central to ClickStack. Deal terms and standalone cloud status remain unresolved, while the MIT repository and ClickStack product line are active.
YC identifies Michael Shi and Warren Lee as the founders of a four-person Summer 2022 company. Shi had led developer experience at LogDNA, later Mezmo, while Lee had been the first site-reliability engineer at an earlier startup.[2] The public record does not establish when they met or their original division of responsibilities.
Shi began DeploySentinel as a side project by January 2022. Its first product was a Chrome and Firefox recorder that converted browser actions into Playwright or Puppeteer tests.[3][4] By the August 2022 Launch HN, it had become an end-to-end testing service focused on flaky tests and maintenance, a problem Shi had encountered in a prior job.[5]
The pivot preserved the debugging mission while widening its surface. The founders had repeatedly jumped among Bugsnag, LogRocket, and CloudWatch, manually carrying identifiers and timestamps between tools to find one root cause. In May 2023, Shi rebranded the suite as HyperDX so continuous-integration and production debugging could live in one product.[6]
HyperDX chose ClickHouse as its database from the beginning and made open source a distribution principle. Explaining the permissive license, Shi wrote, “We want to make observability accessible.”[2] The corpus preserves one exact founder quote suitable for reproduction, not the two requested by the canonical format. No second founder quote is invented.
HyperDX combined logs, traces, errors, session replay, and early metrics in one investigation interface. OpenTelemetry supplied vendor-neutral collection. ClickHouse supplied columnar search. S3 supplied object storage. Trace and session identifiers correlated signals so an engineer could move from a browser replay to a backend span or matching log without copying timestamps across products.[2]

The user flow started with Lucene-style search or optional SQL, continued through native JSON and live tailing, and turned a useful query into an alert or dashboard. Current capabilities include APM and correlation across OpenTelemetry data.[12]
Read the complete post-mortem, the rebuild playbook, and the exact reasons HyperDX is still worth studying now.