2026년 4월: LangChain 뉴스레터

April means we're officially counting down to Interrupt. We’ve got two more meetups on the agent improvement loop before April officially closes out in New York and San Francisco. RSVP while seats still last!

Product Updates

LangSmith

📐 LangSmith now ships with 30+ evaluator templates covering safety, response quality, trajectory, and multimodal outputs, so you're not starting from scratch every time you need to test an agent. Learn more.

🔔 LangSmith now includes cost alerting. Monitor your agent LLM costs and configure alerts when costs exceed expected thresholds. Learn more.

🌐 Arcade.dev's library of 7,500+ agent-optimized tools is now available directly in LangSmith Fleet. Learn more.

🔐 LangSmith Fleet also now includes RBAC/ABAC for tools, usage tracking, and cost management controls for admins. Learn more.

Deploying Deep Agents

📦 deepagents deploy is a single command that takes your Deep Agents harness—model, instructions, tools, skills, sandboxes—and spins up a production-ready, horizontally scalable server. Unlike Claude Managed Agents, Deep Agents stores memory in standard formats you own and query directly. Learn more.

New conceptual guide: The runtime behind production deep agents

This month’s guide breaks down the runtime requirements behind production-ready deep agents, and how `deepagents deploy` packages those capabilities into something teams can ship.

Read the guide →

Join the follow-up webinar →

‍

Interrupt 2026—Agenda Out Now!

Join us May 13-14 in San Francisco for two packed days of talks, workshops, and lessons from teams shipping in production. Industry leaders, including Harrison Chase, Andrew Ng, Chirantan “CJ” Desai, and Aaron Levie will be there to share their insights on the age of agents.

Harrison will talk about the future of agents, what we're seeing work in production, and you'll get a first look at new products. Hear from the teams building at Cisco, Lyft, LinkedIn, Toyota, Rippling, monday.com, Clay, and more. We sold out last year so get your tickets now!

Speak the Lang

See our favorite use cases, best practices, and customer stories covering the most top-of-mind topics in the AI space.

Your harness, your memory

If you're using a closed harness, especially one behind a proprietary API, you don't own your memory. Lose or give up that memory and you've handed over your competitive moat. That's why we built Deep Agents: open source, model-agnostic, using open standards.

Read the full post →

Human judgment in the agent improvement loop

Your team’s expertise, built over years, is what makes a great agent. Learn how you can direct expert attention to the cases that matter, and translate human feedback into test suites for evals with LangSmith.

Read the full post →

Running subagents in the background

As agents take on longer, more complex tasks, the traditional pattern of inline subagents starts to break down. We shipped async subagents in Deep Agents to fix this.

Read the full post →

Upcoming Events

LangChain Hosted

🗽 (April 29) New York // Meetup: The Agent Improvement Loop
RSVP →

🌉 (April 29) San Francisco // Meetup: The Agent and Code Improvement Loops
RSVP →

🌃 (May 4-5) New York // AI Agent Conference: LangChain + Baseten Reception
RSVP →

🌉 (May 19) San Francisco // Meetup: Deep Agents and Deploying to Production
RSVP →

🗽 (May 20) New York // Meetup: Deep Agents and Deploying to Production
RSVP →

🇨🇦 (May 27) Toronto // Meetup: Deep Agents and Deploying to Production
RSVP →

🍀 (May 27) Boston // Meetup: Deep Agents and Deploying to Production
RSVP →

Partner Hosted

⚙️ (April 29) Plano, TX // AI Agents: From PoC to Production
RSVP →

🤝 Customer & integration highlights

Credit Genie built AskGenie, an AI financial assistant inside their mobile app, using LangGraph and LangSmith’s Insights Agent. Insights Agent revealed 36% of conversations were customer support requests, indicating a gap in their existing support chatbot functionality. Read their story →.

Cisco engineers describe what they call "agentic engineering"—a multi-agent model built on LangSmith and LangGraph that treats agents like a coordinated engineering team. The pilot achieved a 93% reduction in time-to-root-cause and saved over 200 engineering hours in a single month. Read their story →.

How can you follow along with the Lang Latest? Check out the LangChain blog, Changelog, and YouTube channel for more product and content updates.

For any additional questions or feedback, email us at hello@langchain.dev.

‍