LangChain: The Agent Framework

After this lesson, you will be able to: Build agents, chains, and memory-backed conversations with LangChain, connect tools, define custom agents, and manage state.

LangChain is the dominant agent framework, opinionated, batteries-included, well-documented. This lesson covers chains, prompt templates, tools, memory, and the modern AgentExecutor pattern. By the end you'll have a working LangChain agent with custom tools.

Prerequisites:Code-First Agent Fundamentals: Python, ReAct, and Tool Calling

Install + setup

1
1. `pip install langchain langchain-anthropic langchain-community`
2
2. Set ANTHROPIC_API_KEY env var.
3
3. Test: `python -c "from langchain_anthropic import ChatAnthropic; print(ChatAnthropic(model='claude-sonnet-4-6').invoke('hi').content)"`

Your first chain

Prompt template → LLM → output parser. The 'Hello World' of LangChain.

python

from langchain_anthropic import ChatAnthropic
from langchain_core.prompts import ChatPromptTemplate
from langchain_core.output_parsers import StrOutputParser

model = ChatAnthropic(model='claude-sonnet-4-6')
prompt = ChatPromptTemplate.from_messages([
    ('system', 'You are a helpful coding tutor for beginners.'),
    ('user', 'Explain {concept} in 3 sentences.'),
])
chain = prompt | model | StrOutputParser()

print(chain.invoke({'concept': 'recursion'}))

Tool-using agent

Built-in tool + custom tool. AgentExecutor handles the loop.

python

from langchain_anthropic import ChatAnthropic
from langchain.agents import create_tool_calling_agent, AgentExecutor
from langchain_core.prompts import ChatPromptTemplate
from langchain_core.tools import tool

@tool
def get_weather(city: str) -> str:
    """Returns the weather for a given city."""
    return f'72F sunny in {city}'  # stub

@tool
def add_numbers(a: int, b: int) -> int:
    """Adds two numbers."""
    return a + b

tools = [get_weather, add_numbers]
model = ChatAnthropic(model='claude-sonnet-4-6')
prompt = ChatPromptTemplate.from_messages([
    ('system', 'You are a helpful assistant.'),
    ('user', '{input}'),
    ('placeholder', '{agent_scratchpad}'),
])
agent = create_tool_calling_agent(model, tools, prompt)
executor = AgentExecutor(agent=agent, tools=tools, verbose=True)

print(executor.invoke({'input': 'Whats the weather in Tokyo and what is 17 + 234?'}))

Memory in LangChain

For multi-turn: use `RunnableWithMessageHistory` to wrap any chain with chat-history awareness. Backing store can be in-memory, Redis, Postgres, DynamoDB. The pattern is the same, wrap, configure session ID per request.

←Code-First Agent Fundamentals: Python, ReAct, and Tool Calling

Back to Code-First Agents

LangGraph: Stateful Multi-Step Agent Graphs→