​
Anthropic Integration

The TypeScript SDK provides automatic instrumentation for Anthropic’s Claude API using OpenTelemetry.

​
How It Works

When you initialize with instrumentModules: { anthropic: Anthropic }:
  1. The SDK instruments the Anthropic client using OpenTelemetry
  2. Every API call is automatically captured as an Event
  3. Token usage and costs are calculated automatically
  4. Both standard and streaming responses are tracked

​
Setup

import * as lai from 'lucidicai';
import { Anthropic } from '@anthropic-ai/sdk';

// Initialize with manual instrumentation (ESM-friendly)
await lai.init({
  sessionName: 'Claude Demo',
  instrumentModules: { anthropic: Anthropic }
});

const anthropic = new Anthropic({ apiKey: process.env.ANTHROPIC_API_KEY });

// All calls are now automatically tracked!

​
What Gets Captured

For every Anthropic API call:
  • Input: Messages and system prompts
  • Model: Claude model used (claude-3-opus, claude-3-5-sonnet, etc.)
  • Output: Complete response content
  • Token Usage: Input and output token counts
  • Cost: Calculated based on current pricing
  • Timing: Request duration
  • Images: Automatically handled for vision requests

​
Known Limitation: Streaming

Due to a limitation in the @traceloop/instrumentation-anthropic library, streaming responses show as “Response received” instead of the actual content. Workaround for streaming: 
const stream = await anthropic.messages.create({
  model: 'claude-3-haiku-20240307',
  messages: [{ role: 'user', content: 'Tell me a story' }],
  stream: true,
  max_tokens: 1000
});

let fullContent = '';
for await (const event of stream) {
  if (event.type === 'content_block_delta' && event.delta.type === 'text_delta') {
    fullContent += event.delta.text;
    process.stdout.write(event.delta.text);
  }
}

// Manually create event with full content
await lai.createEvent({
  description: 'Claude streaming response',
  result: fullContent,
  model: 'claude-3-haiku-20240307'
});

​
Examples

​
Basic Message

const response = await anthropic.messages.create({
  model: 'claude-3-5-sonnet-20241022',
  max_tokens: 1000,
  messages: [{
    role: 'user',
    content: 'What are the key principles of good software design?'
  }]
});

// Automatically tracked with:
// - Model: claude-3-5-sonnet-20241022
// - Token usage
// - Cost calculation
// - Full response

​
With System Prompt

const response = await anthropic.messages.create({
  model: 'claude-3-opus-20240229',
  system: 'You are an expert TypeScript developer.',
  max_tokens: 2000,
  messages: [{
    role: 'user',
    content: 'How do I implement dependency injection in TypeScript?'
  }]
});

// System prompt is captured in the event

​
Vision Requests

const response = await anthropic.messages.create({
  model: 'claude-3-5-sonnet-20241022',
  max_tokens: 1024,
  messages: [{
    role: 'user',
    content: [
      {
        type: 'text',
        text: 'What is in this image?'
      },
      {
        type: 'image',
        source: {
          type: 'base64',
          media_type: 'image/jpeg',
          data: '/9j/4AAQSkZJRg...'
        }
      }
    ]
  }]
});

// Images are automatically tracked

​
Multi-turn Conversation

const messages = [
  { role: 'user', content: 'What is TypeScript?' },
  { role: 'assistant', content: 'TypeScript is a typed superset of JavaScript...' },
  { role: 'user', content: 'What are its main benefits?' }
];

const response = await anthropic.messages.create({
  model: 'claude-3-5-sonnet-20241022',
  max_tokens: 1500,
  messages: messages
});

// Full conversation context is tracked

​
Minimal Example

import * as lai from 'lucidicai';
import { Anthropic } from '@anthropic-ai/sdk';

// Just these 3 lines for full tracking!
await lai.init({ instrumentModules: { anthropic: Anthropic } });
const anthropic = new Anthropic();
const response = await anthropic.messages.create({
  model: 'claude-3-5-sonnet-20241022',
  messages: [{ role: 'user', content: 'Hello Claude!' }],
  max_tokens: 100
});

// That's it! Check your dashboard for the tracked event

​
Cost Tracking

The SDK includes built-in pricing for all Claude models:
ModelInput CostOutput Cost
Claude 3 Opus$0.015/1K$0.075/1K
Claude 3.5 Sonnet$0.003/1K$0.015/1K
Claude 3 Haiku$0.00025/1K$0.00125/1K
Claude 4$0.032/1K$0.096/1K
Costs are automatically calculated and attached to each event.

​
Advanced Usage

​
Temperature and Other Parameters

const response = await anthropic.messages.create({
  model: 'claude-3-5-sonnet-20241022',
  max_tokens: 2000,
  temperature: 0.7,
  top_p: 0.9,
  messages: [{
    role: 'user',
    content: 'Write a creative story about AI'
  }]
});

// All parameters are tracked

​
Error Handling

try {
  const response = await anthropic.messages.create({
    model: 'claude-3-5-sonnet-20241022',
    max_tokens: 100000, // Too high
    messages: [{ role: 'user', content: 'Hello!' }]
  });
} catch (error) {
  // Errors are tracked as events
  if (error.status === 400) {
    console.error('Invalid request:', error.message);
  }
}

​
Using with Steps

// Create a step for complex workflows
await lai.createStep({
  state: 'Document analysis',
  action: 'Extracting key information',
  goal: 'Summarize technical document'
});

const response = await anthropic.messages.create({
  model: 'claude-3-5-sonnet-20241022',
  system: 'You are a technical document analyzer.',
  max_tokens: 3000,
  messages: [{
    role: 'user',
    content: longDocument
  }]
});

await lai.endStep({ evalScore: 95, evalDescription: 'Successfully extracted key points' });

​
Troubleshooting

​
Events Not Appearing

  1. Check import order - Anthropic must be imported AFTER lai.init()
  2. Verify initialization - Ensure instrumentModules included the Anthropic module

​
Streaming Content Missing

This is a known limitation. Use the manual workaround shown above for critical streaming use cases.

​
Custom Models

For fine-tuned or custom models: 
// The SDK will attempt to match the base model for pricing
const response = await anthropic.messages.create({
  model: 'claude-3-5-sonnet-20241022-custom',
  messages: [{ role: 'user', content: 'Test' }],
  max_tokens: 100
});

​
See Also