How to Connect WhatsApp to Claude AI

Combine whatsmeow-node with the Anthropic SDK to build a WhatsApp chatbot powered by Claude. Messages come in via WhatsApp, get sent to Claude for a response, and the reply goes back to the user — with typing indicators while Claude thinks.

Prerequisites

A paired whatsmeow-node session (How to Pair)
An Anthropic API key (set as ANTHROPIC_API_KEY environment variable)
The Anthropic SDK: npm install @anthropic-ai/sdk

Step 1: Set Up Both Clients

import { createClient } from "@whatsmeow-node/whatsmeow-node";
import Anthropic from "@anthropic-ai/sdk";

const client = createClient({ store: "session.db" });
const anthropic = new Anthropic(); // reads ANTHROPIC_API_KEY from env

const SYSTEM_PROMPT = "You are a helpful WhatsApp assistant. Keep responses concise — under 500 characters when possible, since this is a chat interface.";

Step 2: Handle Incoming Messages

client.on("message", async ({ info, message }) => {
  // Skip own messages to avoid loops
  if (info.isFromMe) return;

  const text =
    (message.conversation as string) ??
    (message.extendedTextMessage as { text?: string } | undefined)?.text;
  if (!text) return;

  // Show typing while Claude thinks
  await client.sendChatPresence(info.chat, "composing");

  // Get response from Claude
  const reply = await askClaude(info.sender, text);

  // Send the reply
  await client.sendMessage(info.chat, { conversation: reply });
});

Step 3: Send to Claude

async function askClaude(userJid: string, userMessage: string): Promise<string> {
  const response = await anthropic.messages.create({
    model: "claude-sonnet-4-20250514",
    max_tokens: 1024,
    system: SYSTEM_PROMPT,
    messages: [{ role: "user", content: userMessage }],
  });

  const block = response.content[0];
  return block.type === "text" ? block.text : "I couldn't generate a response.";
}

Step 4: Add Conversation History

For multi-turn conversations, store message history per user:

type Message = { role: "user" | "assistant"; content: string };
const conversations = new Map<string, Message[]>();
const MAX_HISTORY = 20;

async function askClaude(userJid: string, userMessage: string): Promise<string> {
  // Get or create conversation history
  const history = conversations.get(userJid) ?? [];
  history.push({ role: "user", content: userMessage });

  // Trim to last N messages to control token usage
  if (history.length > MAX_HISTORY) {
    history.splice(0, history.length - MAX_HISTORY);
  }

  const response = await anthropic.messages.create({
    model: "claude-sonnet-4-20250514",
    max_tokens: 1024,
    system: SYSTEM_PROMPT,
    messages: history,
  });

  const block = response.content[0];
  const reply = block.type === "text" ? block.text : "I couldn't generate a response.";

  // Store the assistant's reply
  history.push({ role: "assistant", content: reply });
  conversations.set(userJid, history);

  return reply;
}

Complete Example

import { createClient } from "@whatsmeow-node/whatsmeow-node";
import Anthropic from "@anthropic-ai/sdk";

const client = createClient({ store: "session.db" });
const anthropic = new Anthropic();

const SYSTEM_PROMPT =
  "You are a helpful WhatsApp assistant. Keep responses concise — under 500 characters when possible, since this is a chat interface.";

type Message = { role: "user" | "assistant"; content: string };
const conversations = new Map<string, Message[]>();
const MAX_HISTORY = 20;

async function askClaude(userJid: string, userMessage: string): Promise<string> {
  const history = conversations.get(userJid) ?? [];
  history.push({ role: "user", content: userMessage });

  if (history.length > MAX_HISTORY) {
    history.splice(0, history.length - MAX_HISTORY);
  }

  const response = await anthropic.messages.create({
    model: "claude-sonnet-4-20250514",
    max_tokens: 1024,
    system: SYSTEM_PROMPT,
    messages: history,
  });

  const block = response.content[0];
  const reply = block.type === "text" ? block.text : "I couldn't generate a response.";

  history.push({ role: "assistant", content: reply });
  conversations.set(userJid, history);

  return reply;
}

client.on("message", async ({ info, message }) => {
  if (info.isFromMe) return;

  const text =
    (message.conversation as string) ??
    (message.extendedTextMessage as { text?: string } | undefined)?.text;
  if (!text) return;

  console.log(`${info.pushName}: ${text}`);

  // Show typing while Claude thinks
  await client.sendChatPresence(info.chat, "composing");

  try {
    const reply = await askClaude(info.sender, text);
    await client.sendMessage(info.chat, { conversation: reply });
    console.log(`→ ${reply.slice(0, 80)}...`);
  } catch (err) {
    console.error("Claude API error:", err);
    await client.sendMessage(info.chat, {
      conversation: "Sorry, I'm having trouble right now. Try again in a moment.",
    });
  }
});

client.on("logged_out", ({ reason }) => {
  console.error(`Logged out: ${reason}`);
  client.close();
  process.exit(1);
});

async function main() {
  const { jid } = await client.init();
  if (!jid) {
    console.error("Not paired! See: How to Pair WhatsApp");
    process.exit(1);
  }
  await client.connect();
  await client.sendPresence("available");
  console.log("AI bot is online!");

  process.on("SIGINT", async () => {
    await client.sendPresence("unavailable");
    await client.disconnect();
    client.close();
    process.exit(0);
  });
}

main().catch(console.error);

Enhancements

Here are a few ideas to extend this further:

Image understanding — download incoming images with downloadAny(), read the file, and send to Claude as a base64 image block
Rate limiting — track message counts per user to prevent API abuse
Reset command — add a !reset command to clear conversation history
Persistent history — store conversations in a database instead of in-memory for persistence across restarts

Common Pitfalls

Echo loops

Always check info.isFromMe first. Without this, the bot sends a message, sees its own message, sends it to Claude, and replies again — forever.

API key exposure

Never hardcode your API key in the source. Use environment variables (ANTHROPIC_API_KEY) or a .env file (added to .gitignore).

Rate limits on both sides

Both the Anthropic API and WhatsApp have rate limits. For the Anthropic API, handle 429 errors with exponential backoff. For WhatsApp, avoid sending too many messages too quickly — see Rate Limiting.

Prerequisites​

Step 1: Set Up Both Clients​

Step 2: Handle Incoming Messages​

Step 3: Send to Claude​

Step 4: Add Conversation History​

Complete Example​

Enhancements​

Common Pitfalls​

Related Guides

Connect to ChatGPT

Connect to Gemini

Connect to Ollama

Build a Bot