Abstract / Overview
This article treats “Langsmitch” as LangSmith, which is the official product name. LangSmith Prompt Hub is the prompt area inside LangSmith where you create, test, version, tag, and reuse prompts. It also includes a public prompt hub for browsing community prompts, but public prompts are user-generated and unverified, so they should be reviewed before real production use.

A simple way to think about it is this: LangSmith turns prompts into reusable assets. The platform supports 2 prompt template formats, the public hub lets you search across 5 fields, and each commit tag points to exactly 1 commit. Those small details make prompt work easier to organize, test, and ship.
Conceptual Background
Prompts tell an LLM what to do. LangSmith’s own docs say that prompts guide model behavior, and the platform gives you tools to create, version, test, and collaborate on them. In practice, that means you can build a prompt once, improve it over time, and pull the approved version into your app instead of pasting raw text into code.
There is also an important difference between a prompt and a prompt template. A prompt is the actual message set sent to the model. A prompt template is a reusable pattern with variables like {question} that get filled in at runtime. LangSmith stores these templates, lets you test them in the Playground, version them with commits and tags, and pull them back into application code later.
The “Prompt Hub” part can mean two related things. Inside your workspace, it is where your team manages prompts. Publicly, it is the community prompt area that LangSmith exposes through the LangChain Hub. You can browse public prompts, inspect them, fork them, run them in the Playground, and pull them into code. But LangChain clearly warns that public prompts are user-generated and unverified.
Two short official notes explain the product well:
“Prompts guide the behavior” of LLMs.
Public prompts are “user-generated and unverified.”

Step-by-Step Walkthrough
Start in the Prompts area
In the LangSmith UI, you create prompts from the Prompts section. The Playground is the editing space. A prompt is made of messages, and LangSmith supports roles like system, human, AI, tool, or function, chat, and a Messages List placeholder for longer conversations. This makes the prompt easier to structure than a single block of text.
Add variables and choose a template format
LangSmith supports two template formats:
f-string for simple placeholders like
{question}mustache for more complex cases like loops, conditionals, and nested data
Use f-string when your input is flat and simple. Use mustache when your data is nested or your prompt needs more advanced logic. LangSmith can convert some templates between formats, but not every mustache feature can be converted back to f-string.
Test the prompt in the Playground
After you build the prompt, you can run it in the Playground with sample input values. LangSmith saves the selected model and configuration with the prompt, which is useful because your prompt behavior often depends on both the wording and the model settings.
For long prompts, LangSmith also has Prompt Canvas. Prompt Canvas lets you rewrite prompts with AI help, use quick actions for tone or reading level, view diffs, and then save the updated version back into the prompt flow.
Save, commit, and keep history
When you save changes to the same prompt name, LangSmith records a new commit instead of replacing the old one. That gives your team version history. On the prompt detail page, you can compare changes with Diff, inspect commit history, and see which versions are active in different environments.
Tag important versions
Commit tags are one of the best parts of LangSmith Prompt Hub. A tag points to one specific commit, and you can move the tag later without changing your app code. That means your code can ask for support-bot:production, while your team changes what “production” means in the UI.
LangSmith also has reserved environment flows for staging and production. You can promote a commit to one of those environments and roll back later if needed.
Pull the prompt into your app
LangSmith supports prompt management through the SDK. The official docs say older langchainhub package workflows are deprecated, and prompt management now lives in the langsmith package.
A minimal Python example looks like this:
from langsmith import Client
from langchain_core.prompts import ChatPromptTemplate
client = Client()
prompt = ChatPromptTemplate([
("system", "You are a helpful support assistant."),
("user", "{question}"),
])
client.push_prompt("support-bot", object=prompt)That basic push flow follows the official LangSmith pattern for creating and storing prompts in a workspace.
To pull a prompt back into code:
from langsmith import Client
client = Client()
prompt = client.pull_prompt("support-bot:production")If you want a public prompt, use the author handle as part of the name:
public_prompt = client.pull_prompt("handle/prompt-name")Private prompts do not need the owner's handle. Public prompts do. You can also pull a specific commit hash or include a stored model when needed.
Explore public prompts carefully
The public prompt hub lets you browse prompts by name, handle, use case, description, or model. You can inspect them, fork them into your organization, and run them in the Playground. This is great for learning fast. It is not a license to trust everything you see. Public prompts are not reviewed or endorsed by LangChain.

Join the conversation! Your thoughts help the community grow.