LangChain AI Updates: June 30, 2026

1. LangChain Introduces Dynamic Subagents in Deep Agents

LangChain. LangChain introduced dynamic subagents in its Deep Agents framework, a feature that lets an agent write orchestration code rather than make sequential tool calls. The approach has the model write a short script that drives subagent execution, enabling programmatic patterns such as loops, branching, and parallel operations for more reliable handling of large-scale tasks. The post was authored by Sydney Runkle, Colin Francis, and Hunter Lovell. Source

2. LangChain Details How Candidly Built State-Aware Agent Harnesses with LangSmith

LangChain. LangChain published a case study describing how Candidly built state-aware agent harnesses using LangSmith. Authored by Ben Levine and Patrick Hendershott, the piece walks through the company’s approach to managing agent state and harness design on LangChain’s observability tooling. It serves as a practical example of production agent engineering with the LangSmith platform. Source