> For the complete documentation index, see [llms.txt](https://docs.dapta.ai/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.dapta.ai/ai-text-agents/inbox.md). # Inbox Find, assign, answer, release, and close Text Agent conversations in Inbox. Inbox is the operating view, not a complete export of all activity. It opens on **Open** conversations, so business-initiated chats awaiting a reply are not in the default list — add **Awaiting first reply** in the **Status** filter, or clear the filters, to find them. Use history and exports for reporting. Open **Text Agents → Inbox** in the correct workspace. Read the status and assignment before replying: a conversation can be visible without being editable by you. The top of the list has three rows: the **Inbox** title with your availability and the alerts bell; search and the **New conversation** button; then the filter button and the **All**, **My conversations** and **Unassigned** views. To get a desktop alert when a chat is assigned to you or a customer replies on WhatsApp, see [Desktop alerts](/ai-text-agents/inbox/desktop-alerts.md). ## Find the right conversation Use search and combine the available filters, which open from the filter button: **Source**, **Assigned to**, **Status**, date range, **Agents**, and **Tags**. Source distinguishes website widget and WhatsApp activity. **Status** takes several values at once (**Open**, **Awaiting first reply** and **Closed**) and starts on **Open**. Each active filter appears as a chip above the list; click a chip to remove that filter. * **All** is the normal conversation view. Because **Status** starts on **Open**, it does not show chats awaiting their first customer reply until you add that status. * **My conversations** limits the list to your assignments. * **Unassigned** helps find open human-handover conversations without an assigned person; it is not a list of every AI-managed chat. While it is selected, it takes over the **Status** filter, which is locked. * **Unread** can be combined with the assignment view. For example, select your conversations and then unread to focus on your own queue. Opening a conversation clears its unread indicator. Unread is not a delivery, success, or assignment status. When a conversation seems missing, use **Clear filters**: it removes every filter, **Status** included, returns to **All** and turns off **Unread**, so the list shows every conversation, including those awaiting the first reply. Results also depend on your workspace access. ## Understand conversation states | State | Meaning and available work | | -------------------- | ---------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | Open | Active conversation. Assignment and your permissions determine whether you can send manually. | | Awaiting first reply | Your business initiated contact, but the customer has not replied. Read-only: no composer, assignment, tags, close menu, or additional template send within this view. | | Closed | Finished conversation. Review its history; it is not an active free-form reply session. | | Deleted record | A retained deleted record, when surfaced, is read-only; it is not an active chat. | An outbound template does not prove engagement. When the first customer reply arrives, the awaiting conversation becomes open. If no reply arrives before its waiting period expires, it closes; history can report a failed outcome for an unanswered initiation. That path does not run the normal post-conversation analysis. Do not equate “Failed” here with a provider delivery failure. See [history and exports](/ai-text-agents/analytics-and-logs.md) for finding awaiting/unanswered records outside Inbox. ## Take over, assign, and return to the AI 1. Open an **Open** conversation and review the recent messages. 2. Choose yourself or a teammate in **Assign to**. Assignment puts the conversation into human handling. 3. To reply yourself, make sure you are the assigned person and have write access. Assignment to someone else does not enable your composer. 4. When the human work is done but the conversation should continue with the AI, use the **×** control beside the assignment. This removes the assignment and re-enables AI handling; it is different from closing the conversation. 5. Confirm the assignment changed before leaving the chat. Next to the **Inbox** title, the **Online** / **Offline** switch sets your availability in every human-handover pool you belong to at once. It reads **Online** only when you're online in all of them. If you aren't in any pool it's greyed out; ask an admin to add you. Availability does not replace assigning a particular conversation. ## Reply with text or media For an open conversation assigned to you, type your message and send it. **Enter** sends; **Shift + Enter** adds a line break. On WhatsApp, a reply can have up to 4,096 characters, and a counter appears above the message box once the message gets long. A long reply shows a hint such as **Will be sent as 2 messages**: it arrives as consecutive messages, in order, with no numbering such as (1/3). The AI's long replies arrive the same way. Over the limit, the composer shows **Over the 4096-character limit. Shorten it to send.** and won't send it. On supported WhatsApp channels, use the attachment controls to add an image/file or record audio. Allow microphone access when recording, stop and review the recording, then send. Remove unwanted attachments before sending and follow the type/size validation shown by the interface. If an attachment fails, inspect its error before retrying. Do not resend already successful text or attachments just because another item failed. In the timeline you can review: * Customer, AI, and human messages. * Images, downloadable files, and audio playback; a transcription may appear when available. * Tool-call cards with the action name, execution state, parameters, and response. * Messages and media synced from a WhatsApp Business App line that keeps working on the phone. A synchronized Business App message is not an additional AI reply. An action marked completed also does not by itself establish a successful business outcome; read the action result. ## Templates and the reply window If WhatsApp blocks free-form sending, follow the [template guide](/ai-text-agents/channels/message-templates.md). Its customer-service-window rules apply to human replies as well as AI replies. Use **New conversation** for a supported manual template send. In an existing conversation, **Send template** appears only where supported by its provider/state. Awaiting-first-reply and deleted records do not offer that action. Do not repeatedly send opening templates while waiting for the customer. ## Close a conversation Use the conversation's **⋮ → Close conversation**, review the confirmation, and confirm only when the work is finished. Closing is not “return to AI,” and the confirmation warns that the close action cannot be undone. The agent's configured closure/analysis behavior determines subsequent processing for an ordinary conversation. Review its history afterward when you need the final outcome rather than assuming it is immediately available. ## When you cannot send | Situation | What to check | | ------------------------------ | ------------------------------------------------------------------------------------------- | | Composer missing | Conversation state, assignment to you, and write permission. History views are read-only. | | **Message too long** | The reply is over 4,096 characters. Shorten it and send it again; your draft is kept. | | WhatsApp window expired | Use the approved-template workflow; wait for a customer reply before free-form messaging. | | Awaiting first reply | Read-only by design. Monitor for the reply rather than trying to take over. | | Line unavailable or send fails | Check the original sending line in Channels and the provider error. | | Conversation missing | Workspace, access, date/source/assignment filters, and awaiting status; then check history. | *Source review: October 21, 2026. Available actions depend on provider, state, and permissions.* {% content-ref url="/pages/RyyjwoDIy0fMQZ7hWc7V" %} [Desktop alerts](/ai-text-agents/inbox/desktop-alerts.md) {% endcontent-ref %} {% content-ref url="/pages/LT6vPXywEfXjmz1MwdWI" %} [WhatsApp message templates](/ai-text-agents/channels/message-templates.md) {% endcontent-ref %} --- # Agent Instructions This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com. ## Querying This Documentation If you need additional information that is not directly available in this page, you can query the documentation by asking a question. Perform an HTTP GET request on the following URL with the `ask` and `goal` query parameters: ``` GET https://docs.dapta.ai/ai-text-agents/inbox.md?ask=&goal= ``` `ask` is the immediate question: it should be specific, self-contained, and written in natural language. `goal` is what the user is ultimately trying to achieve, the reason they need the answer. Sharing it helps GitBook give you a better, more relevant answer. A goal is most helpful when it describes the outcome the user wants rather than restating the question. For example, with `ask=how do I create an API token`, a goal like `automate deployments from our CI pipeline` lets GitBook tailor the answer to that use case. The response will contain a direct answer to the question and relevant excerpts and sources from the documentation. Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.