FULL STORY

LangChain's Deep Agents: From Open Source to Managed Beta

After Stripe built an AI assistant using LangChain's open-source framework, LangChain launched Managed Deep Agents in public beta to streamline the deployment of production-grade agents.

2026-08-04 ~ 2026-08-11 · 3 episodes · 27 posts

Episode 1 · Stripe Built Internal AI Agent Kai in One Week with Single Engineer (2026-08-04, 4 posts)

Stripe leveraged LangChain's Deep Agents framework to build its internal AI assistant, Kai, in just one week with only one engineer. The tool has achieved massive internal adoption, with 83% of employees using it weekly.

Episode 2 · LangChain Launches Managed Deep Agents in Public Beta for One-Command Deployment (2026-08-08, 19 posts)

On Aug 8, LangChain announced the public beta of Managed Deep Agents (MDA), aiming to provide the simplest path for building, running, and deploying production-grade agents. Founder Harrison Chase noted that agents are increasingly becoming an infrastructure problem, and the product connects the key components needed for large-scale deployment, helping developers transition from prototype to production.

Confirmed

  • Public beta: Managed Deep Agents is now in public beta.
  • Development and deployment: Developers can write agents in Python or TypeScript, test locally, and deploy with a single command to LangSmith-managed runtime.
  • Core advantage: Developers do not need to manage underlying infrastructure, significantly lowering the barrier for production-grade agent deployment.
  • Underlying architecture: Built on the Deep Agents Harness.

Unconfirmed

  • Specific pricing, availability regions, and performance benchmarks have not been announced.

Why it matters

  • Paradigm shift: Harrison Chase divides the evolution of agent building into stages: early frameworks and applications (e.g., LangChain) from late 2022 to early 2023, the subsequent rise of complex architectures, and the current managed agent stage. He believes agent development is shifting from complex underlying architecture to one-click managed services.
  • Infrastructure as a core: Chase emphasizes that future companies will be built around core agent architectures, but productizing them (e.g., persistence, auth, streaming, sandboxing, memory, evaluation) is extremely difficult. A standard agent tech stack is emerging, and managed solutions package these components to address production pain points.
  • Key insight: Chase argues that agents are essentially code, and what truly matters is data and evaluation environments. Building efficient agent systems requires grasping trends like self-improvement, and Managed Deep Agents provides the infrastructure for this goal.

Episode 3 · LangChain Launches Managed Deep Agents (2026-08-09, 4 posts)

LangChain introduced Managed Deep Agents, a highly cost-effective framework designed to simplify the building and deployment of production-grade AI agents. Additionally, Stripe shared details of its internal AI assistant Kai, which was built by one engineer in a week and is now used by 83% of employees weekly.