Skip to main content

​
Overview

Anthropic Claude integration provides access to Claude 3.5 Sonnet, Claude 3 Opus, Haiku, and other models through the Messages API. Supports advanced features like extended thinking, tool use, and vision.

​
Setup

1

Get API Key

Sign up at Anthropic Console and create an API key
2

Add Credential

In Flowise, go to Credentials and create a new Anthropic API credential
3

Add to Canvas

Drag the ChatAnthropic node from the Chat Models category to your canvas

​
Configuration

​
Basic Parameters

​
credential
credential
required
Your Anthropic API credential containing the API key
​
modelName
string
default:"claude-3-haiku"
The Claude model to use:
  • claude-3-5-sonnet-latest - Most capable model
  • claude-3-5-haiku-latest - Fast and cost-effective
  • claude-3-opus-latest - Previous generation flagship
  • claude-3-sonnet - Balanced performance
  • claude-3-haiku - Fastest responses
​
temperature
number
Controls randomness in responses (0.0 to 1.0). Lower = more focused, higher = more creative
​
streaming
boolean
default:true
Enable streaming for real-time response generation

​
Advanced Parameters

​
maxTokensToSample
number
Maximum number of tokens to generate. Claude supports up to 4096 output tokens
​
topP
number
Nucleus sampling threshold (0.0 to 1.0). Use this OR temperature, not both
​
topK
number
Sample from the top K options for each token. Reduces variety of responses

​
Extended Thinking (Reasoning)

Extended thinking is available for Claude 3.5 Sonnet and Claude 4 models, enabling deeper reasoning capabilities.
​
extendedThinking
boolean
default:false
Enable extended thinking mode for reasoning-intensive tasks
​
budgetTokens
number
default:1024
Maximum tokens Claude can use for internal reasoning process (1-10000)

​
Vision Support

​
allowImageUploads
boolean
default:false
Enable image analysis for vision-capable Claude models

​
Usage Examples

​
Basic Configuration

// Standard chat configuration
Model Name: claude-3-5-sonnet-latest
Temperature: 0.7
Max Tokens: 2048
Streaming: true

​
Extended Thinking Mode

// For complex reasoning tasks
Model Name: claude-3-5-sonnet-latest
Extended Thinking: true
Budget Tokens: 4096
Temperature: 0.3  // Lower for more focused reasoning

​
Vision-Enabled Chat

// Analyze images
Model Name: claude-3-5-sonnet-latest
Allow Image Uploads: true
Max Tokens: 2048

​
Tool Use Agent

// Optimal settings for function calling
Model Name: claude-3-5-sonnet-latest
Temperature: 0.1
Top K: 1
Max Tokens: 4096

​
Extended Thinking Deep Dive

When extended thinking is enabled, Claude uses a separate “thinking” process before generating the final response. 
{
  "thinking": {
    "process": "Internal reasoning steps...",
    "tokens_used": 512
  },
  "response": "Final answer based on thinking..."
}
Budget Tokens Guidelines:
  • 1024-2048: Quick reasoning tasks
  • 2048-4096: Moderate complexity problems
  • 4096-8192: Complex analysis and planning
  • 8192+: Advanced research and multi-step reasoning
Extended thinking tokens count against your total API usage but provide transparency into Claude’s reasoning process.

​
Best Practices

Model Selection

  • Use Sonnet 3.5 for most production workloads
  • Use Haiku 3.5 for high-volume, simple tasks
  • Use Opus when you need maximum capability

Extended Thinking

  • Enable for math, coding, and analysis
  • Set budget tokens based on task complexity
  • Monitor token usage in responses

Performance

  • Enable streaming for better UX
  • Use appropriate max tokens
  • Consider caching for repeated queries

Cost Optimization

  • Start with Haiku for testing
  • Use Sonnet for production
  • Monitor extended thinking token usage

​
Claude vs GPT Comparison

FeatureClaude 3.5 SonnetGPT-4o
Context Window200K tokens128K tokens
Output Tokens4K-8K4K-16K
Extended ThinkingYesReasoning (o1 models)
VisionYesYes
Tool UseYesFunction Calling
Best ForAnalysis, writingCreative tasks, coding

​
Common Issues

Anthropic has different rate limits per tier:
  • Free tier: 5 requests/minute
  • Build tier: 50 requests/minute
  • Scale tier: Custom limits
Solution: Implement exponential backoff and monitor usage
Extended thinking requires:
  • Claude 3.5 Sonnet or Claude 4
  • Extended thinking parameter enabled
  • Sufficient budget tokens allocated
Claude supports large contexts but has limits:
  • Monitor conversation length
  • Implement summarization for long chats
  • Use appropriate models (200K context)

Build docs developers (and LLMs) love

Get started for freeTalk to us