LangChain AI Updates: July 30, 2026

1. LangChain Releases Deep Agents v0.7

LangChain released version 0.7 of Deep Agents, its framework for building long-running agents that sustain multi-step tasks. The update focuses on improvements to the underlying agent architecture for durable, extended workflows rather than single-shot calls. It continues LangChain’s push to give developers primitives for agents that plan, delegate, and operate over long horizons. Source

2. Similarweb Details Long-Form Agent Report Evaluation With LangSmith

LangChain published a case study describing how Similarweb uses LangSmith’s evaluation tooling to grade long-form research reports produced by agents. The writeup covers how the team scores comprehensive, multi-section outputs where correctness is spread across a document rather than a single answer. For teams building research agents, it offers a concrete approach to evaluating outputs that do not reduce to a simple pass or fail. Source