Claude API Node.js Example (2026)

Anthropic ships a first-party TypeScript SDK that handles authentication, streaming, and retry logic. The recipes below take you from npm install to a working tool-using agent in under 50 lines. Same code works against api.anthropic.com or any compatible proxy by changing one constructor option.

Install

npm install @anthropic-ai/sdk
# or
pnpm add @anthropic-ai/sdk

Requires Node.js 18+. The SDK is pure ESM with TypeScript types built in.

1. Basic call

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

const client = new Anthropic({
  apiKey: process.env.ANTHROPIC_API_KEY!,
  // baseURL: "https://tokenprovider.store",  // uncomment to use a proxy
});

const msg = await client.messages.create({
  model: "claude-sonnet-4-20250514",
  max_tokens: 256,
  messages: [{ role: "user", content: "What is the capital of Japan?" }],
});

console.log(msg.content[0].type === "text" ? msg.content[0].text : "");

2. Streaming tokens as they arrive

const stream = client.messages.stream({
  model: "claude-sonnet-4-20250514",
  max_tokens: 1024,
  messages: [{ role: "user", content: "Write a haiku about TypeScript." }],
});

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

const final = await stream.finalMessage();
console.log("\nUsage:", final.usage);

The SDK also exposes higher-level helpers like stream.on("text", ...) if you prefer event emitters.

3. Tool use (function calling)

const tools: Anthropic.Tool[] = [{
  name: "get_weather",
  description: "Get current weather for a city",
  input_schema: {
    type: "object",
    properties: {
      city: { type: "string" },
      unit: { type: "string", enum: ["c", "f"] },
    },
    required: ["city"],
  },
}];

const result = await client.messages.create({
  model: "claude-sonnet-4-20250514",
  max_tokens: 512,
  tools,
  messages: [{ role: "user", content: "Weather in Tokyo today?" }],
});

const toolBlock = result.content.find(
  (b): b is Anthropic.ToolUseBlock => b.type === "tool_use",
);

if (toolBlock) {
  console.log("Tool:", toolBlock.name, "Args:", toolBlock.input);
  // …call your real weather function, then send a follow-up:
  // role: "user", content: [{ type: "tool_result", tool_use_id, content }]
}

4. Vision (image input)

import { readFileSync } from "node:fs";

const imageData = readFileSync("./screenshot.png").toString("base64");

const result = await client.messages.create({
  model: "claude-sonnet-4-20250514",
  max_tokens: 512,
  messages: [{
    role: "user",
    content: [
      { type: "image", source: { type: "base64", media_type: "image/png", data: imageData } },
      { type: "text", text: "What does this UI mockup show?" },
    ],
  }],
});

5. Error handling and retries

try {
  const msg = await client.messages.create({ /* … */ });
} catch (err) {
  if (err instanceof Anthropic.APIError) {
    console.error(err.status, err.message);
    // err.status: 401 / 429 / 500 etc.
  } else {
    throw err;
  }
}

// SDK already retries idempotent failures up to 2x with backoff.
// Configure: new Anthropic({ apiKey, maxRetries: 3 })

6. Streaming + tools combined

Real agents stream the assistant's reasoning text and detect tool calls as they appear in the stream. The SDK's stream.on("contentBlock", ...) event fires for each completed block, so you can process tool calls without buffering the entire response.

Switching to a proxy

const client = new Anthropic({
  apiKey: process.env.ANTHROPIC_API_KEY!,
  baseURL: "https://tokenprovider.store",   // only line to change
});

The proxy implements the full Messages API contract — same SDK methods, same response shapes, same streaming events. Use the same key for Claude Code, Cursor, and your scripts.