LangChain

Improving an agent systematically requires a feedback loop. This loop is powered by traces. https://lnkd.in/ecdfrhV7 Traces capture the full execution of an agent run: ✅ Every LLM call ✅ Every tool invocation ✅ Every retrieval step ✅ Every intermediate output ✅ The sequence of decisions that connected them A raw trace tells you what happened. An enriched trace, scored by evaluators and annotated by reviewers, tells you what to do about it. Read our blog by Head of Product Sam Crowder to get in the loop on all things traces, and see how LangSmith connects every step of the process.

Clay is coming to Interrupt! At Interrupt, the Agent Conference by LangChain, Head of AI Jeff Barg will take the stage to show us how Clay successfully scales their GTM Engineering agents. Catch Jeff’s talk along with all the others at Interrupt on May 13-14 in San Francisco. Get tickets here 👉 interrupt.langchain.com

If you're taking an agent to production, you need guardrails. Deep Agents lets you add guardrails to production agents via middleware, which lets you hook into any point in the agent loop. Two common use cases 1. Control flow: model/tool retries, fallbacks when API calls go awry 2. Business logic: PII redaction, compliance checks, custom routing See the full list of LangChain's built-in middlewares here: https://lnkd.in/exyqxVBH. You can also write custom middleware!

Until today, Deep Agents shipped with a single set of prompts, tools, and middleware aimed to work well across all Large Language Models. With the launch of harness profiles, you can now control these parameters on a per-model basis.

“Tracing in LangSmith feels like going from basic psychology to neuroimaging for our agentic AI; we were able to see what was actually happening inside the brain of our system.” https://lnkd.in/ewyy9CUK Madrigal Pharmaceuticals' multi-agentic platform started with a question. “How do we integrate, search, and synthesize information from diverse datasets at scale?” In weeks instead of months, their team was able to build a system using LangChain and LangSmith that can search across sources, reason over structured criteria, synthesize outputs at scale, and operate autonomously. Read their full case study to learn: ✅ How their parallel research sub-agents work under an orchestrator to produce trusted answers whenever a research question is asked. ✅ How their skills framework allows domain experts to be part of development and for a speedier feedback loop. ✅ How tracing in LangSmith allowed them to gain visibility into agent behavior for the first time. Every decision, traceable back to the source.

LangChain is entering a strategic partnership with Axtria - Ingenious Insights to help life sciences enterprises turn agentic experiments into operational results in production. Axtria's pharma-native AgentOps framework, built on LangSmith, gives life sciences enterprises the traceability, compliance controls, and evaluation rigor to move agents from pilot to production with confidence. Our joint solution delivers: • Agents as a Service: Pre-built, pharma-validated agents across Commercial, Medical Affairs, Patient Services, and Data Engineering deployable and governed from day one, without building from scratch. • Pharma-Native Observability & Governance: Enterprise-grade visibility, compliance enforcement, and audit-readiness built specifically for regulated environments covering GxP traceability, immutable audit trails, and model version tracking. • Cost Intelligence: Portfolio-level cost governance that helps enterprises optimize AI spend, right-size model selection, and prevent runaway costs without compromising safety or quality. Regulated environments demand governance that generic tooling wasn’t built to provide. Our joint partnership with Axtria will close this gap.

LangChain is coming to the AI Agent Conference in New York May 4-5, and we will be posted up at Booth 403. Come by and say hello to Charles Bernoskie & me, as we want to know what you're building. We're also hosting a happy hour on Monday night with our friends Baseten, link to sign up is here: https://luma.com/8psixajl

At Listen Labs, agents analyze data from conversations by building their own tables — adding columns like "user sentiment," filling in values automatically, then charting the results. Co-Founder & CTO Florian Juengermann runs through how this works on the Max Agency podcast hosted by Harrison Chase. 🎧 Watch the full episode: YouTube: https://lnkd.in/ede5s6sT Apple Podcasts: https://lnkd.in/e8mPJ3iE Spotify: https://lnkd.in/eF56StZS

Meet our Interrupt MC: Jake Broekhuizen. https://lnkd.in/eQKSDp2J Jake joined LangChain as one of our first deployed engineers, helping leading companies connect data & context to LLMs and build enterprise agents that work in the real world. In two weeks, Jake will adding MC to his resume, as he takes the stage at our agent conference Interrupt. We sat down for a quick Q&A on of what attendees can be expect, trends in the industry, and his recommendations for anyone visiting San Francisco for the first time.

I had the pleasure of joining Michael Kennedy on the Talk Python to Me podcast a few weeks back to talk about Deep Agents, LangChain's newest open source library. Give this episode a listen to learn about how to build agents that can handle complex tasks with deep agents as the base harness. https://lnkd.in/e362tVhX