LangChain's monthly newsletter rounds up product updates: Managed Deep Agents and LLM Gateway have entered public beta, LangSmith Engine now finds agent issues over 2x better with 25% better proposed fixes, and Tuned Evaluators automate quality feedback on production traces. Bring Your Own Cloud (BYOC) is now available on AWS for LangSmith. On the open source side, Deep Agents v0.7 cuts base input tokens by 65% at comparable performance. The newsletter also covers a new video series, an updated Deep Agents course, upcoming meetups and roadshows, and case studies from Stripe and Apollo building agents on Deep Agents.
Table of contents
LangSmith UpdatesOpen SourceWhat does it mean to "own your intelligence"?Speak the LangUpcoming EventsWhat teams are building with LangChainQuestions this post answers
What changed in Deep Agents v0.7 from LangChain?
Deep Agents v0.7 simplifies the base harness, resulting in 65% fewer base input tokens while maintaining comparable performance to prior versions. This reduces token costs and latency for teams running deep agent workflows without requiring changes to agent behavior or quality. Developers tracking agent framework updates like this follow release notes on daily.dev.
What does LangChain's LLM Gateway do for agents in production?
LLM Gateway sits between agents and the models they call, providing cost controls, rate limits, model fallbacks, and sensitive data handling. It is aimed at the operational controls teams need first once agents move into production, and it is currently available in public beta. Teams evaluating LLM infrastructure options compare tools like this on daily.dev.
What improvements did LangSmith Engine get for finding agent issues?
LangSmith Engine became more than 2x better at finding agent issues, with proposed fixes that perform 25% better on standard benchmarks compared to before. This came alongside new Tuned Evaluators that automatically attach quality feedback to production traces, starting with a Perceived Error evaluator. Engineers debugging production agent behavior track tooling gains like this on daily.dev.