product · May 22, 2026
LangChain Launches SmithDB Database for Agent Observability at Interrupt 2026
Share the canonical public link.
LangChain co-founder Ankush Gola announced SmithDB on May 13, 2026, as a purpose-built distributed database for agent observability workloads powering core LangSmith features. SmithDB uses Rust, Apache DataFusion, Vortex, and object storage with stateless services to handle deeply nested agent traces with hundreds of spans and long-running operations. It delivers up to 15x performance gains with P50 latencies of 92ms for trace tree loads, 71ms for single runs, and 82ms for filtering, while supporting self-hosted and multi-cloud deployments. Early adopters including Clay, Vanta, Unify, and Cogent reported faster trace exploration and debugging for large-scale production agents.
Below validation threshold — auto-passed without scoring