The Lang Universe Map

Lesson 1: The Lang Universe Map

Ask five people what "LangChain" is and you'll get five different answers — because the word now covers a whole family of tools, not one library. This course maps the entire constellation: the framework, the orchestration engine, the observability platforms, and the deployment layer. By the end you'll know what each piece is for, how they slot together, and — just as important — when to swap one out for something else.

Mental model: think of the Lang universe as four layers of a stack — build (LangChain + LCEL), orchestrate (LangGraph), observe & evaluate (LangSmith or Langfuse), and serve (LangServe / LangGraph Platform). They're layers, not competitors — except where two tools occupy the same layer (that's LangSmith vs. Langfuse).
THE LANG UNIVERSE — FOUR LAYERS SERVE LangServe · LangGraph Platform (managed or self-hosted) OBSERVE & EVALUATE LangSmith (SaaS) · Langfuse (open source, self-hostable) · Phoenix, Weave, Helicone… ORCHESTRATE LangGraph — state machines, agents, checkpoints, human-in-the-loop BUILD LangChain + LCEL — models, prompts, parsers, retrievers, tools

Where it all came from

LangChain launched in late 2022, right as ChatGPT made LLM APIs mainstream. Its original pitch: give developers composable building blocks — model wrappers, prompt templates, output parsers, retrievers — so you don't hand-roll the plumbing for every app. Harrison Chase founded LangChain Inc. (now also known as LangChain) around it, and the ecosystem grew outward from that core: an expression language for composing chains (LCEL), a graph engine for stateful agents (LangGraph), an observability platform (LangSmith), and a serving layer (LangServe). Langfuse came from outside the company — an independent, open-source observability project that competes with LangSmith on the observe layer.

PieceLayerMakerLicense / model
LangChainBuildLangChain Inc.Open source (MIT)
LCELBuildLangChain Inc.Part of LangChain
LangGraphOrchestrateLangChain Inc.Open source (MIT)
LangSmithObserve & evalLangChain Inc.Proprietary SaaS (no self-hosting)
LangfuseObserve & evalIndependent OSS (team now part of ClickHouse)Open source (Apache 2.0) + paid cloud
LangServeServeLangChain Inc.Open source — archived May 2026, use Platform/FastAPI
LangGraph PlatformServeLangChain Inc.Commercial (managed) / self-hosted
LangMemBuild (memory)LangChain Inc.Open source (MIT)
Not every "Lang" thing is LangChain's. Langfuse is a separate company. There are also unaffiliated projects (Langtrace, Langroid, Langfun) that share the naming vibe. Always check the repo owner before assuming a "lang-*" tool is part of the official ecosystem.

How they work together

A typical production app touches every layer: you build your pipeline with LangChain components and LCEL; for anything agentic you hand the control flow to LangGraph; you wire LangSmith or Langfuse in with a couple of lines of config to trace every call, run evals, and catch regressions; and you ship it behind LangServe or the LangGraph Platform's API server. The observability layer is the only spot with genuine overlap — and choosing between LangSmith and Langfuse is the decision most teams actually agonize over.

Quick takeaway: If you only remember one sentence — LangChain builds, LangGraph orchestrates, LangSmith/Langfuse observes, LangServe/Platform ships.

🧠 Knowledge Check

1. Which layer does LangGraph occupy in the stack?

2. LangSmith and Langfuse occupy the same layer. What's the key difference?

3. Which of these is NOT made by LangChain Inc.?

Further Reading