Log inStart Free Trial
Meta Tech Provider

Automate WhatsApp without code: ask your AI or use the editor

You can automate a WhatsApp Business number without code in two ways. Describe the automation to an AI assistant connected to the number, such as ChatGPT or Claude, and it builds and tests it for you; or pick a starting point in a visual editor and set the trigger and the replies yourself. Both create the same automation, and both can test it on a sample message first.

easyCoexistence connects the WhatsApp Business number you keep in the app to Meta's Cloud API through Coexistence and lets you automate it without code: ask ChatGPT or Claude through the connector, with more than 30 tools, or build in the editor.

11 starting points

New automation gallery templates: Agent answers everything, Menu with options, Capture leads, Welcome message, Away message, Keyword reply, Start from scratch, and more.

3 triggers

Number automation triggers: A message arrives (every message, keyword, first message, after hours), A contact gets a tag, Your system calls a URL.

Up to 60 steps

After its trigger, an automation holds up to 30 steps per line, 60 in all, with branches splitting by rules or your AI.

On by default

New automations start switched On, in the editor and through create_automation; save with the switch Off to test first.

More than 30 tools

What the MCP connector offers, among them create_automation, test_automation and update_automation, which build, check and switch automations.

24 hours

Meta's customer service window: replies that are not templates go out within 24 hours of the customer's last message.

What do you build when you automate WhatsApp without code?

An automation has two parts, and you choose both from menus or describe them in words. The trigger decides what starts it: for an incoming message, Every message, Has keyword (up to 30 keywords, any one is enough), First message, or After hours with your timezone, open days and hours; with no message, A contact gets a tag, or Your system calls a URL. The steps decide what goes back, in order, up to 30 per line and 60 in all, and they can branch by rules or by your AI. The grid shows six of them; the editor also offers questions with buttons, pictures and files, AI agents, contact fields, tags and calls to your own API. Neither mode asks for code, JSON or a server: an assistant fills in the same fields the editor shows you.

Reply with textA fixed message of up to 4,096 characters, with {{first_name}}, {{name}}, {{message}} or {{phone}} filled in
Send a templateOne of the number's approved templates, with its body variables
Reply with AIYour model writes the reply from your Instructions, on your own Anthropic, OpenAI or OpenRouter key
Call my webhookHands the message to n8n or your backend and sends back the reply it returns
WaitPauses from 1 minute to 23 hours before the next step
StopEnds the automation; allowed only as the last step

Should you ask your AI or use the editor?

Pick by where you already work. Owners who spend the day in ChatGPT or Claude describe the result and let the assistant fill in the fields; owners who like to see the whole flow use the diagram. Both write to the same automation, so one built by asking can be opened and changed in the editor, and the other way round.

Ask your AIBuild it yourself in the editor
WhereA chat in ChatGPT or Claude with the easyCoexistence connectorThe Automations tab of the dashboard
What you doDescribe what should happen, in your own words and languagePick a starting point, then edit the trigger and each step
Set up onceAdd the connector and sign in; no key to pasteNothing beyond your login
TestingThe assistant calls test_automation and reports the replyTest opens Try it; Run test shows the result step by step
Going liveAsk it to switch the automation on (update_automation)Set the switch to On and click Save
Plan neededChatGPT Business, Enterprise or Edu for tools that write; Claude from the Free planOnly your easyCoexistence plan

How do you build an automation by asking your AI?

First add easyCoexistence to the assistant, once. In ChatGPT on the web, if easyCoexistence appears in the Plugins directory for your workspace, add it from there; otherwise open Plugins, click + and choose Add custom MCP server, paste https://easycoexistence.com/api/mcp and sign in. In Claude, open Customize > Connectors, click Add, choose Add custom connector, paste the same URL, then Connect and sign in. Sign-in runs through OAuth with your easyCoexistence account, so no key or password goes into the chat.

Then ask for the result, not the settings. Say in plain words when it should run ("when someone asks for our address"), what goes back, and that you want it switched off until you read the test. The assistant finds the number, builds the automation, runs a test and asks before switching it on:

A request in ChatGPT or Claude, and the calls it makes
You: When someone asks for our address or location on WhatsApp, reply with
     "We're at 12 Baker St, open 8:00 to 18:00: https://maps.example/bloom".
     Create it switched off and test it with "where are you?".

→ list_numbers
→ create_automation {
    "connection_id": "…from list_numbers…",
    "name": "Address",
    "enabled": false,
    "trigger": {
      "type": "message_received",
      "match": "keyword",
      "keywords": ["address", "location", "where are you", "endereço", "dirección"],
      "keywordMode": "contains"
    },
    "steps": [{ "type": "reply_text",
      "text": "Hi {{first_name}}! We're at 12 Baker St, open 8:00 to 18:00: https://maps.example/bloom" }]
  }
→ test_automation { "automation_id": "…", "text": "where are you?" }

AI:  Built and tested. The test reply reads "Hi Test! We're at 12 Baker St…". Switch it on?
You: Yes.
→ update_automation { "automation_id": "…", "enabled": true }

How do you build an automation in the editor?

Open the dashboard, go to the Automations tab and click New automation. A gallery offers examples such as Agent answers everything, Menu with options, Send your menu or catalog, Capture leads, Welcome message, Away message, Keyword reply and Send to my system, each with a Use this button, plus Start from scratch. The automation opens as a diagram: the trigger block on top, the steps below it in order. Select a block to edit it in the side panel: Which messages and the keywords on the trigger, the Message box on Reply with text, Model and Instructions on Reply with AI, Webhook URL on Call my webhook. Use + (Add a step here) between two blocks to insert a step and choose its kind. Under Advanced, set how long the automation pauses on a chat you answer from the phone, 4 hours by default, and how many AI replies one chat can get per hour, 10 by default. The labels appear in English, Portuguese or Spanish.

How do you check an automation before customers see it?

New automations start switched On, so saving one puts it to work at once. To check first, set the switch to Off, click Save, then click Test: the Try it panel runs the saved steps on a sample message you type. Nothing is sent to anyone, but the AI step and your webhook really run, so the result shows the actual reply, step by step, with the tokens an AI step used. An assistant does the same through test_automation.

Two things a test cannot show. A Send a template step is not checked with Meta during a test, so confirm that the template's status is APPROVED. And a test ignores the pause after you answer a chat from your phone, which real messages from that chat respect. When the result reads right, switch it On and click Save; from then on the Runs tab keeps 30 days of real runs, with the reason for any skip.

How do you change an automation later?

The same two ways. In the editor, open the automation, change any block and click Save; switching it Off pauses it without losing anything, and Delete removes it together with its run history. Through your assistant, ask for the change in words: it calls list_automations to find the automation and update_automation to apply the change. Fields you do not mention keep their value. A new list of steps replaces the old list, though, and new settings replace both settings at once, so when you change the pause hours, restate the AI limit in the same request.

Where does a no-code automation hand over to other tools?

The common jobs, welcome, away, keyword answers and AI replies, fit entirely in the editor or a chat. These are the points where another tool takes part:

  • Branches: Split into paths separates cases by a contact field, a tag, the message text or opening hours, and AI picks the path chooses by meaning. A menu is one automation with a question, a split and a way back.
  • Live data such as order status, stock or free time slots comes from your backend through Call an API, whose answer later steps can use, or from an n8n workflow through Call my webhook: low-code rather than no-code.
  • Messages your business starts, such as reminders, begin with a tag (A contact gets a tag) or a call from your own system (Your system calls a URL), which then sends an approved template. There is no schedule trigger.
  • Voice notes are written out when you have an OpenAI or OpenRouter key saved, and reach automations as "[voice message]" followed by the words; without one, only "[voice message]".

How to set it up

  1. Connect your WhatsApp Business numberConnect it to easyCoexistence through Meta's Coexistence flow. The app stays on the phone with its chats, and numbers connected without a webhook URL start with automations on.
  2. Save an AI key, if the automation writes with AIOn the Automations page, click AI keys, add an Anthropic, OpenAI or OpenRouter key and click Test and save. Fixed replies, templates and webhooks need no key.
  3. Describe it, or open New automationAsk your assistant for the automation switched off, or pick a starting point in the editor, set the switch to Off and click Save.
  4. Run a testAsk the assistant to test it, or click Test and then Run test with a message a customer would really send. Read the reply as that customer would.
  5. Switch it OnTurn it on through the assistant, or with the switch and Save, and look at the Runs tab after the first real messages.

Common mistakes

  • Saying "create" and expecting a draft. create_automation turns the automation on unless told otherwise; ask for it switched off, read the test, then ask for it on.
  • Pasting an AI provider key into ChatGPT or Claude. No tool accepts a key; save it under AI keys on the Automations page, where it is checked and stored encrypted.
  • Building on a number set to Straight to your system. Its automations do not run; select Receive through easyCoexistence on the number's page, and add your old webhook URL as the Forward URL to keep receiving.
  • Changing one setting through the assistant and losing the other. Settings sent with update_automation replace both values, so a custom AI limit returns to 10 unless the request restates it.
Doing this with EasyCoexistence

Fixed replies, templates, AI replies on your own key, waits and webhook hand-offs are tested before they go live, with no extra charge for automations. 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

Do I need a paid ChatGPT or Claude plan to build automations by asking?

For ChatGPT, a custom MCP server with tools that write needs a Business, Enterprise or Edu plan on the web; Pro connects custom servers with read access only. Claude allows one custom connector on the Free plan and more on paid plans. The editor needs neither, since it is part of the dashboard.

Can I describe the automation in Portuguese or Spanish?

Yes. The assistant takes the request in any language it understands and fills in the same fields, and the replies can be written in your customers' language. The editor itself shows its labels in English, Portuguese and Spanish; other languages see English.

Does building automations with an assistant cost tokens?

Building runs inside your ChatGPT or Claude plan. Tokens are billed only when a Reply with AI step writes, by your provider on your own key, and testing that step spends tokens too, because the model really runs. easyCoexistence adds no charge for automations or runs.

Can an automation built by asking be edited in the editor?

Yes. There is one automation either way, listed in the Automations tab with its trigger and steps. Open it to change any block, or ask your assistant later; list_automations shows the assistant the current version before it edits.

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