LangSmith Unveils Engine v2 to Enhance Safety and Control in AI Agents
Published on: September 26, 2026
LangSmith, a company specializing in tools for autonomous AI development, has launched Engine v2 alongside enhancements to its Managed Deep Agents platform. These updates focus on bolstering safety, oversight, and operational flexibility in agentic AI systems.
Engine v2 introduces automated validation of remedial actions following issue detection, and includes expanded red‑teaming capabilities. These features are designed to help developers more rigorously test agents and ensure that corrective steps are properly verified.
Managed Deep Agents now support identity‑scoped authentication, meaning individual users can be uniquely identified and authenticated. The platform also adds user‑level memory, which allows agents to retain context specific to a given user, and built‑in web search functionality to enable real‑time information retrieval.
Additional enhancements include session navigation tools that allow tracking and revisiting agent interactions over time, fine‑tuning capabilities for adapting agents to specialized tasks, and customizable interfaces for interacting with agent data. These improvements aim to streamline both development and oversight workflows.
These updates come amid increased focus on AI safety as autonomous agents become more capable and widespread. By embedding validation, authentication, memory, and monitoring features directly into the platform, LangSmith is responding to growing demand for trustworthy and controlled AI systems.
No comments yet.