LangChain announced a suite of new products at Interrupt 2026 targeting the full agent development lifecycle. LangSmith Engine autonomously watches production traces, clusters failures, and opens PRs with fixes. SmithDB is a new Rust-based database built on Apache DataFusion delivering up to 15x faster performance for agent observability workloads. Managed Deep Agents provides an API-first hosted runtime for long-running agents with durable execution and sandboxed code execution. LangSmith Sandboxes are now generally available, offering hardware-virtualized microVM environments for safe code execution. Context Hub centralizes versioning and management of agent instructions and policies. LangSmith LLM Gateway adds a governance layer with spend limits and PII redaction. Fleet gained sandbox access, five prebuilt agents, and free model usage. Deep Agents 0.6 adds a code interpreter REPL, typed streaming, and delta-based checkpoint storage. LangChain Labs was announced as an applied research effort on continual learning from agent traces.
