Capability · Framework — orchestration
LangChain Hub
LangChain Hub is the 'npm for prompts' of the LangChain ecosystem. Prompts live as versioned, taggable objects inside LangSmith and can be pulled into code with a single API call. Public-hub contributors share common patterns (ReAct, self-ask, RAG); private workspaces host team prompts behind auth. Hub + LangSmith tracing is one of the most polished prompt-lifecycle workflows in use today.
Framework facts
- Category
- orchestration
- Language
- Python / TypeScript
- License
- Hosted SaaS (clients are MIT)
Install
pip install langchain langsmith
export LANGSMITH_API_KEY=lsv2-...Quickstart
from langchain import hub
from langchain_anthropic import ChatAnthropic
# pull a well-known public prompt
prompt = hub.pull('hwchase17/react')
chain = prompt | ChatAnthropic(model='claude-opus-4-7')
print(chain.invoke({'input':'What is MCP?','agent_scratchpad':'','tools':'','tool_names':''}).content)Alternatives
- Humanloop — prompt management with non-dev editors
- Braintrust — versioned prompts + evals
- PromptLayer — prompt registry alternative
- Local prompts in git — the zero-dependency baseline
Frequently asked questions
Is LangChain Hub open-source?
No — the registry is hosted inside LangSmith. The client APIs are open-source as part of the LangChain packages.
Can I use Hub without LangChain?
You can pull prompt JSON via REST or the LangSmith SDK and use the text however you like. But `hub.pull` is most useful with LangChain's Runnables because it returns instantiated prompt objects.
Sources
- LangChain Hub docs — accessed 2026-04-20
- LangSmith — accessed 2026-04-20