💬Messages API Deep Dive

Master conversations, system prompts, and response handling

Messages API Deep Dive

The Messages API is the foundation of everything you build with Claude.

Anatomy of a Request

JavaScript
import Anthropic from "@anthropic-ai/sdk";

const client = new Anthropic();

const response = await client.messages.create({
  // REQUIRED
  model: "claude-sonnet-4-6",           // Which model
  max_tokens: 1024,                      // Max output length
  messages: [                            // Conversation
    { role: "user", content: "Hello!" }
  ],

  // OPTIONAL
  system: "You are a helpful teacher.",  // System prompt
  temperature: 0.7,                      // Creativity (0-1)
  stop_sequences: ["END"],               // Custom stop words
});

System Prompts — Setting Claude's Behavior

System prompts act as persistent instructions:

JavaScript
// A customer support bot
const response = await client.messages.create({
  model: "claude-sonnet-4-6",
  max_tokens: 1024,
  system: `You are a customer support agent for TechCo.
Rules:
- Be friendly and professional
- If you don't know something, say so
- Never share internal policies
- Always offer to escalate to a human agent`,
  messages: [
    { role: "user", content: "My order hasn't arrived yet" }
  ]
});

Multi-Turn Conversations

Build up conversations over time:

JavaScript
let conversationHistory = [];

async function chat(userMessage) {
  conversationHistory.push({
    role: "user",
    content: userMessage
  });

  const response = await client.messages.create({
    model: "claude-sonnet-4-6",
    max_tokens: 1024,
    messages: conversationHistory
  });

  const assistantMessage = response.content[0].text;

  conversationHistory.push({
    role: "assistant",
    content: assistantMessage
  });

  return assistantMessage;
}

// Usage
await chat("What is JavaScript?");
await chat("How is it different from Python?");  // Claude remembers context!

Temperature: Controlling Creativity

Temperature	Effect	Good For
0	Deterministic, same answer each time	Math, facts, code
0.3	Mostly consistent, slight variation	Customer support
0.7	Balanced (default)	General conversation
1.0	Most creative and varied	Creative writing, brainstorming

Handling Responses Properly

JavaScript
const response = await client.messages.create({...});

// Get the text
const text = response.content[0].text;

// Check why it stopped
if (response.stop_reason === "max_tokens") {
  console.log("Response was cut off — increase max_tokens");
}

// Check token usage
console.log(`Input: ${response.usage.input_tokens} tokens`);
console.log(`Output: ${response.usage.output_tokens} tokens`);

Streaming Responses

Get responses token-by-token (great for UIs):

JavaScript
const stream = await client.messages.stream({
  model: "claude-sonnet-4-6",
  max_tokens: 1024,
  messages: [{ role: "user", content: "Write a poem" }]
});

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

Next up: Text generation — summaries, content, code, and more.

Module 5

2/7

📋 Understanding JSON & Data Formats

Text Generation Mastery ✍️

2/7