n8n WhatsApp Chatbot Cost: Meta's New Fees, Free JSON

Since October 1, 2026, Meta charges for WhatsApp service replies (free-form messages you send inside the 24-hour customer service window) after the first 1,000 per business phone number each month. The rate is $0.0034 per delivered message to US and Canadian numbers and $0.0113 to Australian numbers (A$0.0148 on an AUD account). Add n8n’s execution fee and model tokens, and a bot handling 500 five-message conversations a month costs about $57 on n8n Cloud Pro, or about $16 self-hosted: 3 to 11 cents per conversation for North American customers.

The cost figures are a worked example from list prices, with every assumption shown below. Rates checked 29 Oct 2026 on Meta’s developer pricing page and WhatsApp’s rate calculator; n8n prices from n8n.io (US view). CAD at the Bank of Canada rate of 1.4036 (28 Oct 2026), AUD at the RBA rate of 0.7053 USD per AUD (29 Oct 2026). Prices exclude tax.

What changed on October 1, 2026

From Meta’s pricing page:

• Service messages are charged again. They had been free since November 1, 2024. Every business phone number now gets a free tier of 1,000 delivered service messages per month; Meta charges from the 1,001st. Unused messages don’t roll over.
• Utility templates sent inside an open customer service window are charged again. They had been free since July 1, 2025.
• No payment method, no paid service messages. Without one, Meta delivers service messages within the free tier but stops after it.
• Government and eligible non-profit accounts keep free service messages beyond the free tier through December 31, 2027.

Meta also says it may change pricing only on the first day of a quarter (January 1, April 1, July 1, October 1), with at least one month’s notice for rate changes. The next possible change is January 1, 2027.

One caution: the marketing copy on WhatsApp’s own pricing site still said “we do not charge for service messages” when we checked, while its rate calculator showed a service rate. Meta’s developer documentation is the one that describes the October 1 change, so we follow it.

WhatsApp Cloud API pricing per message: US, Canada and Australia

Meta charges per delivered message, by category and by the recipient’s country calling code. The US and Canada (+1) share the North America region. Australia (+61) is in Rest of Asia Pacific.

Source: the WhatsApp message rate calculator, read 29 Oct 2026. Meta has no CAD rate card; Canadian businesses bill in USD or another supported currency, and you can’t change currency on an existing Messaging account without Meta’s migration process. Utility and authentication messages also have volume tiers that lower the rate at high monthly volumes.

Free entry point windows. If a user messages you from a Click-to-WhatsApp ad (on Android or iOS) and you reply within the service window, Meta opens a free window of up to 7 days in which marketing, utility, authentication and service messages aren’t charged.

Can you send WhatsApp marketing messages to US numbers?

No. Meta’s documentation says: “WhatsApp does not currently deliver marketing template messages to WhatsApp users with United States phone numbers” (Meta). Attempts return an error. Canadian numbers aren’t covered by that block. If your bot is for US customers, plan around service replies and properly categorized utility templates (order updates, appointment reminders), not promotions.

How an n8n WhatsApp chatbot uses executions

n8n counts executions by trigger type. For webhook triggers, it counts “one execution for every inbound request that activates the trigger. This includes requests with an empty body” (n8n docs). WhatsApp delivers each incoming message to your workflow as a webhook, so:

• 1 customer message = 1 execution.
• Your bot’s replies happen inside that execution (they cost Meta fees, not executions).
• Unverified for your setup: if your WhatsApp app or trigger is also subscribed to message status updates (sent, delivered, read), each of those may arrive as its own request and start an execution. That could double or triple your count. Check your execution list in the first week, and filter or unsubscribe from statuses you don’t use.

The plan math: 500 conversations × 5 customer messages = 2,500 executions, exactly Cloud Starter’s allowance with zero headroom. Pro (10,000) is the realistic floor.

Free n8n WhatsApp AI chatbot template (importable JSON)

WhatsApp Trigger → IF (is it a text message?) → AI Agent with memory keyed on the sender’s WhatsApp ID → WhatsApp Send.

{
  "name": "WhatsApp AI chatbot (unboxth starter)",
  "nodes": [
    {
      "parameters": { "updates": ["messages"] },
      "name": "WhatsApp Trigger",
      "type": "n8n-nodes-base.whatsAppTrigger",
      "typeVersion": 1,
      "position": [0, 0]
    },
    {
      "parameters": {
        "conditions": {
          "conditions": [
            { "leftValue": "={{ $json.messages?.[0]?.type }}", "rightValue": "text", "operator": { "type": "string", "operation": "equals" } }
          ]
        }
      },
      "name": "Is text message?",
      "type": "n8n-nodes-base.if",
      "typeVersion": 2,
      "position": [220, 0]
    },
    {
      "parameters": {
        "promptType": "define",
        "text": "={{ $('WhatsApp Trigger').item.json.messages[0].text.body }}",
        "options": {
          "systemMessage": "You are the WhatsApp assistant for Example Co. Reply in under 60 words. If the customer asks for a person, say a team member will reply within business hours.",
          "maxIterations": 3
        }
      },
      "name": "AI Agent",
      "type": "@n8n/n8n-nodes-langchain.agent",
      "typeVersion": 1.7,
      "position": [440, -20]
    },
    {
      "parameters": { "model": "claude-haiku-5-5", "options": { "maxTokensToSample": 300 } },
      "name": "Anthropic Chat Model",
      "type": "@n8n/n8n-nodes-langchain.lmChatAnthropic",
      "typeVersion": 1.3,
      "position": [380, 200]
    },
    {
      "parameters": {
        "sessionIdType": "customKey",
        "sessionKey": "={{ $('WhatsApp Trigger').item.json.contacts[0].wa_id }}",
        "contextWindowLength": 5
      },
      "name": "Simple Memory",
      "type": "@n8n/n8n-nodes-langchain.memoryBufferWindow",
      "typeVersion": 1.3,
      "position": [520, 200]
    },
    {
      "parameters": {
        "operation": "send",
        "phoneNumberId": "YOUR_PHONE_NUMBER_ID",
        "recipientPhoneNumber": "={{ $('WhatsApp Trigger').item.json.contacts[0].wa_id }}",
        "textBody": "={{ $json.output }}"
      },
      "name": "Send WhatsApp reply",
      "type": "n8n-nodes-base.whatsApp",
      "typeVersion": 1,
      "position": [700, -20]
    }
  ],
  "connections": {
    "WhatsApp Trigger": { "main": [[{ "node": "Is text message?", "type": "main", "index": 0 }]] },
    "Is text message?": { "main": [[{ "node": "AI Agent", "type": "main", "index": 0 }], []] },
    "AI Agent": { "main": [[{ "node": "Send WhatsApp reply", "type": "main", "index": 0 }]] },
    "Anthropic Chat Model": { "ai_languageModel": [[{ "node": "AI Agent", "type": "ai_languageModel", "index": 0 }]] },
    "Simple Memory": { "ai_memory": [[{ "node": "AI Agent", "type": "ai_memory", "index": 0 }]] }
  }
}

After pasting it onto an n8n canvas, add your WhatsApp Business Cloud credentials to the trigger and send node, replace YOUR_PHONE_NUMBER_ID, and pick Gateway credits or your own Anthropic key on the model node. The trigger subscribes to messages only, which avoids status updates. Node versions and field paths can differ between n8n releases; treat this as a starting point we haven’t tested on every version, and run a few test messages before publishing.

Cost per conversation: worked example at 500 and 5,000 a month

Assumptions (estimated from list prices):

• Each conversation: 5 customer messages and 5 bot replies, all inside the 24-hour window, so replies are service messages.
• One business phone number, so one 1,000-message free tier.
• Model: Claude Haiku 5.5 at about 6,000 input and 500 output tokens per reply, $0.00085 per reply (Anthropic).
• n8n: one execution per customer message; Cloud Pro at $50/mo annual (10,000 executions) for 500 conversations, Pro’s 50,000 tier at $120/mo annual for 5,000. Self-hosted on a Hostinger KVM 2 VPS at $8.79/mo (a 24-month promo that renews at $14.99, from our self-hosting guide).
• No Click-to-WhatsApp free windows, no status-webhook executions.

500 conversations a month (2,500 replies, 1,500 billable after the free tier; LLM $2.13):

On an AUD-billed Meta account, the Australian Meta line is 1,500 × A$0.0148 = A$22.20.

5,000 conversations a month (25,000 replies, 24,000 billable; LLM $21.25):

The pattern: at low volume n8n’s plan is the biggest line, and at high volume Meta’s per-message fee takes over, especially for Australian customers, whose rate is more than 3x North America’s. Whether one small VPS handles 25,000 executions a month depends on your workflow; we haven’t load-tested it.

n8n Cloud vs self-hosted for a WhatsApp bot

• Cloud is simpler: HTTPS webhooks work out of the box, and you can pay for the model through Gateway credits. You pay for headroom you may not use.
• Self-hosted removes the execution fee, but you need a public HTTPS URL for Meta’s webhook, backups, and upgrades. The full trade-off, including upkeep time, is in n8n self-hosted vs cloud cost.

For a broader platform comparison, see n8n vs Make vs Zapier pricing.

Meta Business Agent vs your own n8n agent

Meta’s own AI responder is billed at $2.00 per 1 million tokens, a single rate that covers the AI processing and delivery. Meta says one message typically uses 20,000 to 25,000 tokens, “approximately 4–5 cents (USD) per message”, with no free tier, and it stays charged inside free entry point windows (Meta).

Meta’s agent is less work to set up. Your own agent is cheaper per message and lets you choose the model, the memory and the tools. If you’re also pricing voice, see AI voice agent cost per minute, and for the agent token math in more depth, n8n AI agent cost per run.

How to keep WhatsApp chatbot costs down

• Answer in fewer, longer replies. Meta charges per delivered message, so one complete reply costs less than three short ones.
• Stay inside the 24-hour window. Service replies inside it are the cheapest category, and the window resets with every customer message.
• Use Click-to-WhatsApp ads for campaigns. Replies inside the free entry point window aren’t charged for up to 7 days.
• Categorize templates accurately. Utility costs far less than marketing; follow Meta’s template categorization rules rather than labelling promotions as utility.
• Use a small model for FAQs and hand off to a person or a bigger model only when needed.
• Filter status webhooks so they don’t eat executions.
• Collect consent. You may only message users who have opted in to hear from you. Our n8n lead generation workflow covers consent capture for inbound leads.

Billing notes for Canadian and Australian businesses

• Canada: there is no CAD rate card, so a Canadian business is billed in USD (or another supported currency) and its card issuer converts. At 1.4036, a US/CA service message is about C$0.0048.
• Australia: you can open your Messaging account in AUD and be billed A$0.0148 per service message to Australian numbers, avoiding conversion fees. Currency is chosen when you create the Messaging account; Meta offers a currency migration API to switch later.
• Payment method: add one in Meta Billing Hub before you pass 1,000 service messages in a month, or replies will stop being delivered.

More automation guides are in the AI automation hub.