LangChain/Blog/25/10/Agent Frameworks v1.0

• LangChain and LangGraph Agent Frameworks Reach v1.0 Milestones

  • Metadata

    • Author: Sydney Runkle and the LangChain OSS team
    • Published: October 22, 2024
    • Source: LangChain Blog
    • Annotated in Readwise: Link

  • Key Highlights

    • LangChain 1.0 and LangGraph 1.0 released - first major versions
    • Updated langchain to focus on core agent loop
    • New middleware concept for flexibility
    • Upgraded model integrations with latest content types
    • Redesigned documentation site

  • Framework Purposes

    • LangChain - Fastest way to build an AI agent with standard tool calling architecture, provider agnostic design, and middleware for customization
    • LangGraph - Lower level framework and runtime for highly custom and controllable agents, designed for production-grade, long running agents

  • LangChain 1.0 Improvements

    • create_agent Abstraction

      • Built around the core agent loop for quick startup
      • Built on LangGraph runtime for reliable agents
      • Prebuilt and user defined middleware enable step by step control and customization
      • Similar to create_react_agent from langgraph.prebuilts (used in production for a year)

    • Middleware System

      • Fine grained control at every step an agent takes
      • Built-in middlewares for common use cases:
        • Human-in-the-loop: Pause agent execution for user approval, editing, or rejection of tool calls
        • Summarization: Condense message history when approaching context limits
        • PII redaction: Identify and redact sensitive information before content is passed to the model
      • Support for custom middleware that hooks into various points in the agent loop

    • Structured Output Generation

      • Improved structured output generation in agent loop
      • Reduced latency and cost by eliminating extra LLM call
      • Fine grained control over structured output generation via tool calling or provider-native structured output

    • Standard Content Blocks

      • Provider-agnostic spec for model outputs
      • Consistent content types across providers
      • Support for reasoning traces, citations, and tool calls
      • Typed interfaces for complex response structures
      • Full backward compatibility

  • LangGraph 1.0 Features

    • Durable state - Agent execution state persists automatically
    • Built-in persistence - Save and resume agent workflows at any point
    • Human-in-the-loop patterns - First-class API support for human review
    • First stable major release in durable agent framework space
    • After more than a year of iteration and adoption by companies like Uber, LinkedIn, and Klarna

  • Breaking Changes

    • create_react_agent deprecated in langgraph.prebuilt (moved to langchain.agents)
    • Python 3.9 support dropped (EOL October 2025)
    • v1.0 requires Python 3.10+
    • Python 3.14 support coming soon
    • Package surface area reduced, old functionality moved to langchain-classic

  • When to Use Each Framework

    • Choose LangChain 1.0 for:

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

    • Choose LangGraph 1.0 for:

      • Workflows with mixture of deterministic and agentic components
      • Long running business process automation
      • Sensitive workflows requiring more oversight/human in the loop
      • Highly custom or complex workflows
      • Applications where latency and/or cost need careful control

  • Installation

    • Python:
      • uv pip install --upgrade langchain
      • uv pip install langchain-classic
      • uv pip install --upgrade langgraph
    • JavaScript:
      • npm install @langchain/langchain@latest
      • npm install @langchain/langchain-classic
      • npm install @langchain/langgraph@latest

  • Migration Resources

    • Release overviews: Python, JavaScript
    • Migration guides: Python, JavaScript

  • New Documentation Site

    • Completely redesigned documentation site at docs.langchain.com
    • All LangChain and LangGraph docs—across Python and JavaScript—in one unified site
    • Parallel examples, shared conceptual guides, and consolidated API references
    • More intuitive navigation, thoughtful guides, and in-depth tutorials

  • Community Impact

    • 90M monthly downloads
    • Production applications at Uber, JP Morgan, Blackrock, Cisco, and more
    • Commitment to stability: no breaking changes until 2.0