Log inStart Free Trial
Meta Tech Provider

WhatsApp auto-reply on the number you keep in the app

A WhatsApp auto-reply is a message sent without anyone typing it. The WhatsApp Business app has two, the greeting and away messages, sent from the phone. A number connected to Meta's Cloud API through Coexistence can also auto-reply from a server with fixed text, an approved template or an answer written by your own AI, while you keep the app and step in any time.

easyCoexistence connects your WhatsApp Business number to Meta's Cloud API through Coexistence, so auto-replies run on a server while you keep the app: fixed text, approved templates, or your own AI on your Anthropic, OpenAI or OpenRouter key, with no charge for AI and no extra charge for automations.

4,096 characters

Longest fixed reply a Reply with text step sends, with {{first_name}}, {{name}}, {{message}} and {{phone}} filled in.

20,000 characters

Room for the Instructions of a Reply with AI step: your business facts, tone and rules.

4 hours

Default pause on a chat after you answer it from the phone app; set from 0 to 168 under Advanced.

10 per hour

Default cap on AI replies per chat, adjustable from 1 to 60, so a loop with another bot stops at the cap.

14 days

Inactivity after which the WhatsApp Business app's own greeting message goes to a returning customer again.

No change

What Meta's Coexistence guide lists for the app's greeting and away messages after onboarding; the Cloud API does not support them.

What kinds of WhatsApp auto-reply are there?

Two families, and a Coexistence number can use both, though not for the same job. The WhatsApp Business app's own tools live on the phone: a greeting message for new contacts and an away message on a schedule, each a fixed text, with recipients filtered by your address book if you like. Automations on Meta's Cloud API run on a server, and you choose what starts them and what they send:

WhatsApp Business appCloud API automation
Sends fromThe phone, while it has an internet connectionA server, with the phone off or offline
Starts onA first message or one after 14 days of silence; an away scheduleAny message, a keyword, a first message, after hours, a tag, or a call from your system
Replies withOne fixed textFixed text, buttons, a picture or PDF, an approved template, your own AI, or your webhook
Who gets itEveryone, or filtered by your address bookAnyone whose message matches the trigger
RecordThe chat itselfA run log for 30 days, step by step

How do you set a fixed-text auto-reply?

Choose the trigger for the moment, then write the text once. In the Automations tab, New automation offers Welcome message, which starts on a contact's first message, Away message, which starts after hours, and Keyword reply, which starts on words you choose. Each opens with a Reply with text step whose Message box holds up to 4,096 characters, with {{first_name}}, {{name}}, {{message}} and {{phone}} filled in from the customer's details. Texts that work:

  • Acknowledge: "Hi {{first_name}}! Got your message. We answer from 9 a.m. to 6 p.m., Monday to Friday, usually within the hour."
  • Point to the answer: "Here is our price list: [link]. Tell me what you need and I'll check availability for you."
  • Closed: "Thanks for writing! We're closed now and open again at [9 a.m.]. For urgent orders, call [phone]."
  • Echo the question: "Thanks, {{first_name}}! You asked: {{message}}. A person from our team will reply here shortly."

How do you set an AI auto-reply?

Save your own key, then let the model write each reply from your instructions. Click AI keys on the Automations page, then Add key beside Anthropic, OpenAI or OpenRouter, paste the API key and click Test and save: the key is checked with the provider without spending tokens, saved encrypted, and shown afterwards only as "Added, ending in" its last four characters. Then add a Reply with AI step from the step menu (for example on a Blank automation), and fill in Model, picked from the list your key loads; Instructions, up to 20,000 characters; Earlier messages it reads, 0 to 20 with 10 by default; and Longest reply (tokens), 50 to 2,000 with 600 by default. For longer conversations, the Agent answers everything example starts an AI agent instead, which can also use files, tags and a hand-off.

Under your instructions, each AI reply also carries built-in rules: say a person from the business will answer soon when the customer asks for one or the AI cannot help, never invent prices, stock, dates or policies your instructions leave out, and say it is the business's automated assistant when asked. The first AI reply to a person in a day also opens by saying so.

If the provider later refuses the key, the automation shows "Paused: your OpenAI key was refused. It starts again by itself once you save a key that works.", you get one email, and it is left out of matching until then, so a lower-ranked automation, such as an away message, can answer in its place.

Instructions for a Reply with AI step (example)
You are the WhatsApp assistant of Ana's Bakery.
Hours: Tuesday to Sunday, 8:00 to 19:00. Closed on Mondays.
Delivery: within 5 km, for orders placed before 17:00.
Prices: sourdough 4.50, croissant 1.80, birthday cakes from 25.00.
Answer in two or three short sentences, in the customer's language.
For custom cakes, complaints or refunds, say a person will reply today.

Can your AI set up the auto-reply for you?

Yes. With the easyCoexistence connector added to ChatGPT or Claude, describe the auto-reply in a chat. The AI calls list_numbers to find your number, create_automation with the trigger and the reply, and test_automation to run it on a sample message without sending anything, the AI step running for real on your key, then tells you what it built. Ask it to create the automation switched off when you want to read the test before it goes live, and to switch it on afterwards. A Reply with AI step needs your provider key saved in the dashboard first; no tool accepts a key.

A request in ChatGPT or Claude, and the calls it makes
You: Reply to messages on my WhatsApp number with AI, using my OpenAI key.
     You're Bloom Café's assistant: open 8:00 to 18:00, delivery within 5 km.
     Create it switched off and test it first.

→ list_numbers
→ create_automation {
    "connection_id": "…from list_numbers…",
    "name": "AI assistant",
    "enabled": false,
    "trigger": { "type": "message_received", "match": "any" },
    "steps": [{
      "type": "ai_reply",
      "provider": "openai",
      "model": "…a model your key lists…",
      "prompt": "You are Bloom Café's assistant. Open 8:00 to 18:00. We deliver within 5 km.",
      "history": 10,
      "max_tokens": 600
    }]
  }
→ test_automation { "automation_id": "…", "text": "Do you deliver to Rua Augusta?" }
→ update_automation { "automation_id": "…", "enabled": true }

What happens when you reply from the phone?

The automation steps back on that chat. On a Coexistence number, messages typed in the WhatsApp Business app reach the API as smb_message_echoes events, and an automation treats one as you taking over: for the next 4 hours by default (the same when someone answers from Conversations), runs on that chat are skipped as "You answered this chat from your phone recently". Change it per automation under Advanced, in "Pause on a chat after I answer it (hours)", from 0 to 168; at 0 the automation keeps replying in chats you answer. A reply you send during a Wait step also cancels the rest of that run. The customer sees one conversation either way, since replies the automation sends appear in the app's chat on your phone.

Should you keep the app's greeting and away messages on?

Only for jobs no automation does. Meta's Coexistence guide lists the app's messaging tools, greeting and away messages included, as unchanged after onboarding and not supported on the Cloud API. They keep firing from the phone, and an automation can neither see their settings nor switch them off. If an automation already greets new customers or answers after hours, turn the app's version off in the Tools tab, under Greeting message or Away message, or the customer receives both texts. Keep the app's version where you need what only it does, such as sending only to people outside your address book.

How to set it up

  1. Open AutomationsIn the easyCoexistence dashboard, open the Automations tab and click New automation.
  2. Choose the kind of replyPick Welcome message, Keyword reply or Away message for fixed text, or Agent answers everything for an AI reply; each opens the editor pre-filled.
  3. Write the replySelect the Reply with text step and edit the Message box, or fill in the agent's Instructions, or the Reply with AI step's Model and Instructions.
  4. Test it switched offSwitch the automation Off, click Save, then open Test, type a sample message and click Run test. Nothing is sent to anyone.
  5. Turn it onSwitch it On and click Save; it answers the next message that matches its trigger.

Common mistakes

  • Running a Welcome message and an AI agent together and expecting both on a first message. Only one automation runs per incoming message, so the welcome takes a new customer's first message and the AI answers from the second on.
  • Writing "Hi {{first_name}}!" without reading it back. The variable is the first word of the WhatsApp profile name the customer chose, which can be a nickname, an emoji or empty, so the greeting can read "Hi !".
  • Writing AI instructions without your facts. The model is told not to invent prices, stock, dates or policies missing from the Instructions, so a prompt without them leads to vague replies or hand-offs to a person.
  • Setting Earlier messages it reads to 0 for chats that run over several messages. The model then sees only the latest message and can ask again for what the customer already said.
Doing this with EasyCoexistence

The automation steps back on a chat when you answer it from the phone. US$ 9 down to US$ 0.75 per number per month, falling as you add numbers, with the first 3 days free.

easyCoexistence is a multichannel API and MCP server that connects AI agents such as Claude and ChatGPT to WhatsApp, Instagram, TikTok, Facebook and more. On WhatsApp it uses Coexistence: the business keeps using its number in the WhatsApp Business app while the AI works through Meta's official API.

Frequently asked questions

Can an auto-reply send a picture or a document?

Yes. The Send a picture or file step sends a picture or PDF you uploaded to your files, such as a menu or a price list, with a caption if you like, inside the 24 hours after the customer last wrote. Templates with a header picture or a button variable cannot be sent by automations yet. Your own system can also send media through the Cloud API with the number's credentials.

Do auto-replies work in WhatsApp groups?

No. Meta does not synchronize group chats on a Coexistence number with the Cloud API, so auto-replies run in one-to-one chats. Groups keep working in the app as before.

Can I auto-reply only to new customers?

Yes, with the First message trigger. It runs on the first message a contact sends after automations start receiving the number's messages, and again when someone writes after 30 days without a message from them.

What does an AI auto-reply cost?

Your AI provider bills the tokens to your own key, and easyCoexistence never charges for AI or marks it up. Each reply sends the model your Instructions and the earlier messages you allow, 10 by default and up to 20, and stops at the Longest reply you set, 600 tokens by default. Meta bills messages to your own WhatsApp Business Account after a free tier of 1,000 service messages per number each month.

Keep reading

Ready to get started?

Set up WhatsApp Coexistence in minutes, not months. The app keeps working on the phone.

Start Free Trial

Verified on