Introduction

The modern LLM application stack

Four frameworks cover most of what you need to ship production AI apps — composition, orchestration, observability, and data. This site is a concise, opinionated reference for each.

LangChain

Compose LLM apps from modular building blocks.

Read docs

LangGraph

Build stateful, controllable agents as graphs.

Read docs

LangSmith

Observability and evals for LLM applications.

Read docs

LlamaIndex

The data framework for LLM applications.

Read docs

Haystack

Production-ready pipelines for search and RAG.

Read docs

DSPy

Programming — not prompting — language models.

Read docs

CrewAI

Role-playing multi-agent teams.

Read docs

AutoGen

Conversational multi-agent programming.

Read docs

Pydantic AI

Type-safe agents with Pydantic at the core.

Read docs

Semantic Kernel

Enterprise SDK for LLM orchestration.

Read docs

Not sure where to start?

The comparison page breaks down what each framework is best at and how they fit together.