LangChain

What it does

LangChain is an open source Python framework for building AI agents. You write code that supplies a model, a set of tools (plain functions the agent can call) and a system prompt, and the create_agent function returns an agent you invoke with a list of messages. It gives you a single interface for chat models and embeddings across providers such as OpenAI, Anthropic, Google, Azure, AWS Bedrock, Ollama and Hugging Face.

Developers use it to assemble tool-calling agents, then adjust behavior through middleware, for example adding human-in-the-loop approval, compressing long conversations, or removing sensitive data. Agents run on LangGraph's durable runtime, which provides persistence, checkpointing, rewind and human-in-the-loop support. The framework is meant to let you swap models, tools and databases without rewriting your application. It connects to LangSmith, a separate paid platform for tracing, evaluation and deployment, using environment variables.

It suits software developers and engineering teams comfortable writing Python who want a customizable agent harness rather than a no-code builder. The library is MIT-licensed and free to use, installed with pip. You supply your own model-provider access, which carries its own costs. LangSmith tracing and deployment are optional add-ons with separate pricing. Teams needing heavier orchestration may prefer LangGraph, and the docs point to Deep Agents for a batteries-included option.

What it can do

• Create tool-calling agents with the create_agent function
• Standard interface for chat models and embeddings across providers
• Middleware hooks for human approval, conversation compression and sensitive-data removal
• Durable execution with persistence, checkpointing, rewind and human-in-the-loop
• Optional tracing and evaluation through LangSmith via environment variables

Works with

• OpenAI
• Anthropic
• Google Gemini
• OpenRouter
• Fireworks
• Baseten
• Ollama
• Azure
• AWS Bedrock
• Hugging Face

Example workflows

Build a tool-calling agent

Input: a Python function and a model identifier. Steps: install langchain plus a provider package, define the tool, call create_agent with model, tools and system prompt, then invoke it with a user message. Output: a response message from the agent after any tool calls.

Add approval and tracing to an agent

Input: an existing agent. Steps: add middleware for human-in-the-loop approval, set LANGSMITH_TRACING=true and an API key. Output: an agent that pauses for approval and sends traces to LangSmith for debugging.

Before you choose

• The library is free, but model-provider usage and optional LangSmith services are billed separately.
• Code-first Python framework; non-programmers would need a different tool such as the separate no-code LangSmith Fleet.
• LangChain, LangGraph, Deep Agents and LangSmith are distinct products; choose the layer that fits your needs.

Listed under

• Coding
• DevOps and observability
• Sandboxes and infrastructure
• Agent framework
• Agent observability
• AI agents
• Deployment
• Developers
• LLM

Get started

Install per the docs; model provider access is separate

1. Install langchain and a provider package, for example pip install -qU langchain "langchain[openai]"
2. Obtain an API key for your chosen model provider
3. Follow the Quickstart to define tools and call create_agent

The LangChain library is MIT-licensed and free. Model-provider costs apply. The optional LangSmith platform has its own plans, not covered here.

Source: langchain.com/langchain, checked 29 Oct 2026. Prices can change; check before you buy.

Sources for this overview

Product pages, documentation, and pricing checked 29 Oct 2026. Each source supports part of the overview; availability and prices can change.

• LangChain: Open Source AI Agent Framework for Any Model ↗www.langchain.com
• LangChain overview - Docs by LangChain ↗docs.langchain.com
• LangSmith Plans and Pricing ↗www.langchain.com
• Home - Docs by LangChain ↗docs.langchain.com

Company website

langchain.com: product, pricing and policy pages

YC directory mirror

Batch and company name

Social accounts

Accounts the company links, and its posts on X

Human review

Checked by a person

knowledge_base

Last checked 29 Oct 2026. Every field keeps its source and date; unknown means we couldn’t confirm it.