# LangChain Reaches 1.0 - Whats new?

Source: https://www.youtube.com/watch?v=hjIlj4OAECQ
Recap page: https://rapidrecap.app/video/hjIlj4OAECQ
Generated: 2025-11-12T00:31:17.526+00:00

---
## Quick Overview

LangChain released version 1.0, focusing on three core goals: providing a new create_agent abstraction, building on the LangGraph runtime for reliable agents, and streamlining the surface area by trimming the namespace, while LangGraph 1.0 focuses on lower-level control for custom, production-grade agents requiring features like durable state and human-in-the-loop patterns.

**Key Points:**
- LangChain raised $125M at a $1.25B valuation to build the platform for agent engineering, announced alongside LangChain and LangGraph 1.0 releases.
- LangChain 1.0 introduces a new create_agent abstraction for building agents quickly with any model provider, built on the LangGraph runtime.
- LangGraph 1.0 focuses on lower-level control, providing production-ready features like durable state and human-in-the-loop patterns for highly custom agents.
- LangChain is positioned as the fastest way to build an AI agent with standard patterns, while LangGraph is for highly custom, long-running, production-grade agents.
- LangSmith, the agent engineering platform, now includes Observability, Evaluation, Deployment (formerly LangGraph Platform), and Agent Builder (in private preview).
- LangChain 1.0 addresses past feedback by streamlining its surface area and introducing middleware for fine-grained control over the agent loop.
- The new documentation site was also launched alongside the 1.0 releases for both frameworks.

![Screenshot at 00:18: LangChain announces its Series B funding round: "LangChain raises $125M to build the platform for agent engineering," contextualizing the importance of the 1.0 releases.](https://ss.rapidrecap.app/screens/hjIlj4OAECQ/00-00-18.png)

**Context:** The video discusses the major 1.0 releases for LangChain and LangGraph, frameworks central to building LLM applications. Harrison Chase, the creator of LangChain, details the evolution of the project from simple prompt chaining to sophisticated agent engineering, highlighting the need for more stability, control, and production readiness, which these 1.0 releases aim to address through new abstractions and the integration of LangGraph's execution model.

## Detailed Analysis

The video announces the 1.0 releases for LangChain and LangGraph, coinciding with LangChain raising $125 million at a $1.25 billion valuation to build the platform for agent engineering. LangChain 1.0 focuses on making agent building faster and more accessible by introducing the `create_agent` abstraction, which is built on the LangGraph runtime, offering higher-level abstractions and middleware-based customization. LangGraph 1.0, conversely, is a lower-level framework designed for highly custom, production-grade agents requiring precise control, offering features like durable state and human-in-the-loop patterns. The speaker clarifies that LangChain is best for shipping quickly with standard patterns (Model -> Tools -> Response loop), while LangGraph is for complex, long-running, sensitive workflows needing careful control over latency and cost. LangSmith is positioned as the comprehensive platform for agent engineering, offering observability, evaluation, deployment, and the new Agent Builder (in private preview) to manage the entire lifecycle of agent development, addressing past user feedback about LangChain's overly heavy abstractions and unwieldy surface area.

### LangChain 1.0 Focus

- Shipping quickly with standard agent patterns
- Agents fitting the default loop (model -> tools -> response)
- Middleware-based customization
- Higher-level abstractions over low-level control

### LangGraph 1.0 Focus

- Workflows with deterministic/agentic components mixture
- Long running business process automation
- Sensitive workflows needing oversight/human in the loop
- Highly custom or complex workflows
- Applications requiring careful control over latency/cost

### LangSmith Platform Features

- Observability (tracing/metrics)
- Evaluation (testing/scoring)
- Deployment (formerly LangGraph Platform)
- Agent Builder (private preview for no-code text-to-agent)

### LangChain Evolution & Feedback

- Abstractions were too heavy, package surface grew unwieldy, developers wanted more control over the agent loop, leading to the 1.0 rewrite.

### Agent Workflow Visualization

- A diagram shows the agent loop flow: Request -> before_agent -> before_model (splitting to tools or model_call) -> after_model -> after_agent -> Result, highlighting hooks for middleware.

![Screenshot at 00:00: Harrison Chase announces LangChain as a Python package for building LLM applications via a tweet preview.](https://ss.rapidrecap.app/screens/hjIlj4OAECQ/00-00-00.png)
![Screenshot at 00:18: LangChain announces its Series B funding round: "LangChain raises $125M to build the platform for agent engineering," contextualizing the importance of the 1.0 releases.](https://ss.rapidrecap.app/screens/hjIlj4OAECQ/00-00-18.png)
![Screenshot at 00:55: An animated graphic illustrating the transition from chained prompts to Agentic LLMs, highlighting Memory, Tool Calling, and Structured Outputs.](https://ss.rapidrecap.app/screens/hjIlj4OAECQ/00-00-55.png)
![Screenshot at 02:04: A graphic representing GPT 3.5 capabilities like Tool Calling, Structure Output, and Function Calling, noting that early versions lacked these features.](https://ss.rapidrecap.app/screens/hjIlj4OAECQ/00-02-04.png)
![Screenshot at 03:35: A complex flowchart illustrating the structure of an agent execution flow involving state, agent, action, and end nodes.](https://ss.rapidrecap.app/screens/hjIlj4OAECQ/00-03-35.png)
![Screenshot at 03:50: The LangSmith dashboard showing a list of tracing projects with metrics like Run Count, Error Rate, and Latency.](https://ss.rapidrecap.app/screens/hjIlj4OAECQ/00-03-50.png)
![Screenshot at 05:51: A detailed trace view in LangSmith showing the execution steps \(nodes\) of a 'RetrievalGraph' agent over time.](https://ss.rapidrecap.app/screens/hjIlj4OAECQ/00-05-51.png)
![Screenshot at 06:06: The LangSmith 'Monitor' tab displaying charts for Trace Count, LLM Call Count, Trace Latency, and LLM Latency over time.](https://ss.rapidrecap.app/screens/hjIlj4OAECQ/00-06-06.png)
![Screenshot at 07:31: Four individuals discussing LangChain/LangGraph differences around a small table, with plush toys on the table.](https://ss.rapidrecap.app/screens/hjIlj4OAECQ/00-07-31.png)
![Screenshot at 10:19: A slide listing the three core things LangChain 1.0 focuses on: create\_agent abstraction, LangGraph runtime, and streamlined surface area.](https://ss.rapidrecap.app/screens/hjIlj4OAECQ/00-10-19.png)