> ## Documentation Index
> Fetch the complete documentation index at: https://docs.open.cx/llms.txt
> Use this file to discover all available pages before exploring further.

# WhatsApp with Genesys Cloud and OpenCX

> How WhatsApp conversations work when Genesys Cloud is your contact center: replies, the 24-hour window and templates from Genesys sent from the OpenCX inbox.

WhatsApp customers can reach your Genesys Cloud agents through OpenCX, or through Genesys itself. Which one applies depends on where your WhatsApp number is connected.

| Your WhatsApp number | How it works |
| - | - |
| **Connected to OpenCX** | The AI answers. When it hands off, OpenCX opens a Genesys conversation for the session, and your agent works it in Genesys. Every reply goes out from your number. |
| **Connected to Genesys** | Genesys routes the conversation. The AI answers inside your flow with [native messaging](https://docs.open.cx/integrations/genesys/native-messaging.md). With [every Genesys conversation](https://docs.open.cx/integrations/genesys/every-conversation.md) shown, the session also appears in OpenCX. |

## Your number in OpenCX

What each side sees once the AI hands a WhatsApp session to Genesys:

| When | The customer sees | The Genesys agent sees |
| - | - | - |
| The AI hands off | The AI's last message | A conversation that opens with the customer's latest message, with the summary and transcript in participant data |
| A template went out before the handoff, from a workflow, a campaign, the inbox or the API | The template | The template in the transcript, under who sent it, so they know what the customer is answering |
| The agent replies | The reply as typed, once, with numbered lists and links kept | |
| The agent sends quick-reply options | Reply buttons (up to three, 20 characters each), otherwise a list | The customer's tap, as their message |
| The customer writes | | Each message once, in order. The AI stays quiet. |
| The customer replies to a specific message | | `(Replying to: "…")` above the reply |
| The customer reacts, sends a sticker, or sends something WhatsApp can't show | | `[Reacted 👍]`, `[Sticker]` with the image, or a line saying WhatsApp couldn't show it |
| A teammate or a workflow sends a template while the agent has the session | The template, once | A copy in the conversation, marked as sent from OpenCX |
| A reply can't be delivered (outside the 24-hour window, a number not on WhatsApp, a limit) | Nothing | A failed mark on the reply and a line in `opencx_notes` with the reason. The session gets one internal note. |
| The agent replies after the 24-hour window, with **Re-open WhatsApp chats** set | Your template, once, asking them to reply. The agent's reply follows as soon as they do. | A line in `opencx_notes` saying the customer was asked to reply and the message goes out when they do |
| The agent ends the conversation | Nothing more | The session closes in OpenCX |

Good to know:

* **The 24-hour window** counts from the customer's last message to that number, also in a closed session. A message to your other WhatsApp number doesn't open it.
* **Outside the window**, send a template from OpenCX (see [Messaging rules](https://docs.open.cx/whatsapp/whatsapp-rules.md)), or let OpenCX do it with **Re-open WhatsApp chats** below. Once the customer answers, the window is open again and the agent can reply.
* **Delivery:** Genesys marks each agent reply delivered or failed. Genesys doesn't take read receipts from OpenCX, so agents don't see when the customer read a reply.
* **Reactions, stickers and unsupported messages** reach your agents only while the session is handed off. While the AI has the session, they aren't shown and the AI doesn't answer them.

## Your number in Genesys

* **The AI answers in your flow.** With [native messaging](https://docs.open.cx/integrations/genesys/native-messaging.md), the AI replies before the conversation reaches a queue, and a handoff goes through your flow.
* **In a queue, your agents answer.** Genesys only lets OpenCX send approved templates into an ongoing WhatsApp conversation, so the AI can't reply there. With [every Genesys conversation](https://docs.open.cx/integrations/genesys/every-conversation.md) shown, these sessions start with your agents and carry an internal note that says why the AI isn't answering. With assist mode on, the AI writes drafts for the agent.
* **To reach the customer from OpenCX**, use **Send template** with your WhatsApp templates in Genesys. See [Replying from OpenCX](https://docs.open.cx/integrations/genesys/every-conversation.md#replying-from-opencx). The role OpenCX uses needs `responses:response:view` (see [Permissions](https://docs.open.cx/integrations/genesys/permissions.md)).
* **When a template is refused or none are listed**, see [Fix replies, messages and calls](https://docs.open.cx/integrations/genesys/troubleshooting-replies.md#a-whatsapp-template-from-the-inbox-is-refused-or-none-are-listed).

## Settings

Everything else on this page works once Genesys is connected and live. Quick-reply options follow [Channels](https://docs.open.cx/integrations/genesys/channels.md#agent-replies).

| Setting | What it does | Default |
| - | - | - |
| **Re-open WhatsApp chats** (**Channels** tab) | When a Genesys agent replies after the customer's 24-hour window, OpenCX sends this template once to ask the customer to reply, and sends the agent's reply as soon as they do. Only the agent's latest reply is kept, for 72 hours. If sending it fails, OpenCX tries again with the customer's next message, up to three tries, then tells the agent it wasn't sent. A session erased in the meantime drops it. Pick an approved WhatsApp template that needs no values. | Off: a late reply is refused and the agent is told why |

## Related

> [card] [Channels](https://docs.open.cx/integrations/genesys/channels.md)
>
> How agent replies reach each channel.

> [card] [WhatsApp messaging rules](https://docs.open.cx/whatsapp/whatsapp-rules.md)
>
> The 24-hour window and templates.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.