LangChain AI Updates: June 27, 2026

1. LangChain Adds Prompt Caching to Deep Agents

LangChain published a post by Alex Olsen detailing how its Deep Agents framework applies prompt caching across multiple model providers to cut API costs. The framework automatically uses caching strategies such as explicit cache breakpoints and implicit caching, which the post reports can reduce token costs by 49 to 80 percent without extra configuration. This matters most for long-running agents, where the same prompts are reused repeatedly across conversation turns. Source