What Langfuse does
Langfuse is an open-source “AI engineering platform” focused on tracing, prompt management, evaluations, datasets/experiments, and related observability for LLM and agent applications. It is designed to let teams run an iteration loop that connects production telemetry (hierarchical traces and other recorded signals) to offline/online evaluation workflows, then back to prompt/version changes and experimentation.
Langfuse supports both self-hosted (open source) and hosted cloud deployments. Product-wise, Langfuse positions itself as an integrated toolkit rather than a single feature: teams can ingest structured traces from their applications, attach prompt versions and experiment runs, define evaluations (including LLM-as-a-judge style workflows and code-driven evaluators), and then analyze results via dashboards, metrics, and filtered views. The platform emphasizes vendor neutrality and OpenTelemetry-based instrumentation, and it is built to scale for high-throughput ingestion and fast analytical querying.
Strategically, Langfuse joined ClickHouse in January 2026; per Langfuse, it stays open-source and self-hostable and keeps operating its cloud offering unchanged, while combining Langfuse’s application-layer telemetry with ClickHouse-backed infrastructure to improve performance, reliability, and enterprise readiness.