ChatAnthropic: Reference Guide

What is ChatAnthropic?

ChatAnthropic is the LangChain chat model for Anthropic Claude. It speaks the Anthropic Messages API and exposes the same Runnable interface as ChatOpenAI — invoke, ainvoke, stream, astream, and batch — so you can switch between Claude and other providers without rewriting your chains. It takes a list of BaseMessage objects and returns an AIMessage carrying the response text, token usage, stop reason, and any tool calls.

The class covers the full modern Claude feature set: streaming over server-sent events, tool calling through bind_tools, structured output through with_structured_output, vision inputs for images and PDFs, prompt caching to cut cost on repeated context, and extended thinking on the reasoning-capable models where the model exposes a separate thinking block before its answer. Token usage comes back on response_metadata and usage_metadata, so you can track cost without a callback. Claude has a large context window (200K tokens on current models), which makes it a natural fit for long-document RAG and multi-turn agents.

Configuration is straightforward: set ANTHROPIC_API_KEY in the environment or pass api_key directly, and choose a model such as claude-3-5-sonnet, claude-3-5-haiku, or a newer release. The class lives in the langchain-anthropic package, which you install separately from langchain-core. Pin a dated model string in production rather than a floating alias so a provider-side model update never silently changes your output.

When to Use

You need Claude's superior reasoning or longer context windows. Use ChatAnthropic for complex reasoning tasks.

Use Cases

• • Reasoning-heavy tasks
• • Long context processing
• • Vision understanding
• • Code analysis
• • Content generation
• • Multi-turn reasoning

Key Features

• ✓ All Claude versions
• ✓ Extended thinking
• ✓ Long context (200K)
• ✓ Tool use
• ✓ Vision support
• ✓ Streaming

When NOT to Use

Notes

Import

from langchain_anthropic import ChatAnthropic

Initialization Parameters

Code Examples

from langchain_anthropic import ChatAnthropic
from langchain_core.messages import HumanMessage

model = ChatAnthropic(model='claude-3-5-sonnet-20241022', temperature=0)
result = model.invoke([HumanMessage(content='Explain quantum computing in one sentence')])
print(result.content)

from langchain_anthropic import ChatAnthropic

model = ChatAnthropic(model='claude-3-5-sonnet-20241022')
for chunk in model.stream('Write a haiku about retrieval augmented generation'):
    print(chunk.content, end='', flush=True)

from langchain_anthropic import ChatAnthropic
from pydantic import BaseModel, Field

class Ticket(BaseModel):
    priority: str = Field(description='low, medium, or high')
    summary: str = Field(description='one-line summary')

model = ChatAnthropic(model='claude-3-5-sonnet-20241022')
structured = model.with_structured_output(Ticket)
ticket = structured.invoke('The checkout page returns a 500 on every purchase')
print(ticket.priority, ticket.summary)

Common Mistakes

Alternative Models

Related LangChain References

Browse the full LangChain API reference index to explore more classes, methods, and decorators, or start with the LangChain introduction tutorial for end-to-end context on building with ChatAnthropic and the wider framework.

ChatAnthropic FAQ