LangChain API Reference

Access comprehensive API documentation for LangChain – a framework for building applications with large language models (LLMs). Covers Python and JavaScript SDKs for LangChain, LangGraph, LangSmith, and integrations.

When to Use

Use this skill when:

• Working with LangChain Python or JavaScript code
• Looking up class signatures, method parameters, or return types
• Finding the right LangChain component for a task
• Understanding LangGraph agent workflows
• Integrating LangSmith for tracing and observability

Accessing Documentation

MCP Server (Recommended for IDEs)

Use the MCP server for structured API access:

https://reference.langchain.com/mcp

Available tools:

• search_api: Search for classes, functions, symbols by name or keyword
• get_symbol: Get full documentation for a specific symbol

Markdown Endpoints

All documentation pages are available as markdown:

Index Files

URL Patterns

https://reference.langchain.com/{language}/{package}.md                    # Package overview
https://reference.langchain.com/{language}/{package}/{Symbol}.md           # Symbol documentation
https://reference.langchain.com/{language}/{package}/{Symbol}/{member}.md  # Nested member

Languages: python, javascript, java, go

Decision Guide

Python vs JavaScript

Note: APIs are similar but not identical. Always check language-specific documentation.

Gotchas

1. Package name format differs by language

   • Python: langchain-core (hyphenated)
   • JavaScript: @langchain/core (scoped package)

2. Deprecated patterns – Avoid these legacy approaches:

   • LLMChain → Use createAgent or createDeepAgent instead
   • ConversationChain → Use createAgent with message history
   • Direct AgentExecutor → Use createAgent or createDeepAgent
   • LCEL chains → Prefer createAgent for new projects

3. Symbol paths use dots, not slashes

   • Correct: RunnableSequence.invoke
   • URL becomes: /python/langchain-core/RunnableSequence/invoke.md

4. Version differences – The documentation reflects the latest stable versions. Check the version badge on package pages for exact versions.

Common Tasks

Using a Chat Model

from langchain_openai import ChatOpenAI
model = ChatOpenAI(model="gpt-4")
response = model.invoke("Hello!")

Search: ChatOpenAI, ChatAnthropic, ChatVertexAI

Building an Agent

from deepagent import createAgent

agent = createAgent(model="gpt-4", tools=[...])
response = agent.run("Your task here")

Search: createAgent, createDeepAgent, Agent

RAG (Retrieval-Augmented Generation)

Search: VectorStore, Retriever, create_retrieval_chain

Tools and Skills

Search: Tool, BaseTool, StructuredTool

Best Practices

1. Use createAgent or createDeepAgent for building agent workflows
2. Enable tracing with LangSmith for debugging
3. Prefer specific integrations (e.g., langchain-openai) over generic interfaces
4. Check inheritance – Many classes extend Runnable with shared methods

More Resources

• LangChain Documentation – Tutorials and guides
• LangGraph Documentation – Multi-actor workflows
• LangSmith – Tracing and observability