LangChain

in English: LangChain / LangGraph

professionalwidespread

LangChain is an open-source developer toolkit for building AI applications and agents from language models, data sources, and tools; LangGraph is its process-control counterpart.

In plain terms

A language model on its own is like a powerful engine: it can do many things, but it only becomes a car if wheels, a steering wheel, and brakes are added. LangChain is a construction kit for these parts. It provides ready-made elements so that the model can access corporate documents, search the web, calculate, or remember earlier parts of the conversation.

LangGraph is another tool from the same team and is responsible for the workflow. It allows you to map out what steps an agent can take, in what order, where it must step back, and where it must pause for human approval. If LangChain is the parts box, LangGraph is the assembly diagram.

As a user, you almost never see these, because they are not applications intended for the general public, but tools for developers. It is still worth knowing their names because many customer service chatbots, document searchers, and corporate AI assistants are built with such frameworks.

How it works

The developer selects a language model (the framework handles many providers in a unified way), making it relatively easy to switch later.
It connects to tools: search engine, database, calendar, internal document repository. The model can call these if needed for the task.
In LangGraph, the process is described as a graph: nodes are steps (for example, search, writing a response, verification), and edges show what can follow each step, even backwards or in cycles.
The system saves its state while running, so long tasks can be paused and resumed, and checkpoints can be inserted where a human approves the next step.
The operation is logged and measured (for this purpose, the company offers a service called LangSmith) to reveal where the agent made mistakes or became expensive.

An everyday example

In a web shop's chat window, you ask where your package is. In the background, an agent retrieves your order, queries the courier company's data, and if you request a refund, it pauses until a staff member approves it. Such sequences are often built using LangChain and LangGraph.

Who is affected

developerOne of the most well-known toolkits for building applications or agents based on language models, so it frequently appears in job postings and projects. → Code

entrepreneur, executiveIf your company orders or develops an AI solution, the choice of framework affects costs, maintainability, and how tied you become to a supplier.

jobseeker, new graduateFor AI developer roles, this is often a required or advantageous skill, so it is worth knowing what it entails. → Work & HR

A common misconception

MythLangChain itself is an artificial intelligence that makes the language model smarter.

In realityLangChain is not a model but connective software: it links an existing model to data, tools, and workflows. The quality of the response still depends on the model and on how the system is designed.

Limits and risks

The framework does not eliminate the language model's errors: the agent can still make mistakes or hallucinate, just within more structured frameworks.
It faces much criticism for placing too many layers between the model and the developer, which complicates debugging; it is often unnecessary for simple tasks.
It changes rapidly, so older descriptions and example code quickly become obsolete, requiring rewrites for updates.
If the agent can also use tools (send emails, modify data), the developer must separately ensure security and manage permissions.

Similar, but not the same

Large language model – The language model itself generates the text; LangChain is merely the toolkit built around it, which is useless without a model.
Model Context Protocol – The MCP standard defines how tools connect to a model, while LangChain is a programming library that builds entire applications using such standards.
AI-agent – The agent is the finished, independently acting system, while LangChain and LangGraph are possible tools used to build it.

Related concepts

Broader

Agent-keretrendszer, AI-agent, Open-source software

Narrower

LangGraph, LangSmith

On the edge

Tool use, Retrieval-Augmented Generation, Model Context Protocol, Multi-agent system, Orchestration

A brief history

Harrison Chase launched LangChain as an open-source project in October 2022, and in the wave following the launch of ChatGPT, it quickly became one of the most popular AI development tools. LangGraph appeared in early 2024, specifically designed to control more complex, multi-step agents. In autumn 2025, both reached version 1.0, a major release meant to be stable.

Where things stand

By the end of 2026, LangChain and LangGraph are among the most widely used agent-building frameworks, with a stable major version and enterprise users. They face strong competition from agent development kits from model manufacturers and other open frameworks.

Status: 2026 November

In the news

Appears in 3 Agent Age stories; the curve shows weekly occurrence in the news magazine.

GitHub Security Lab automates fuzzing with an AI agent frameworkSep 24, 2026 · Code
Critical sandbox escape found in DeepSeek's open-source coding agent Harness (CVE-2026-82533)Sep 8, 2026 · Code
Insurers turn to AI for claims and back-office work: MS&AD, Korea, NTT DATA, GuidewireAug 3, 2026 · Money

Also known as

LangChain · LangGraph · Langchain · LangChain keretrendszer · LangChain framework · LangSmith · agent-keretrendszer · agent framework

Sources

This entry was machine-edited and has undergone machine expert review; the “In the news” section updates automatically from Agent Age stories.