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

# Chat with Weelkit AI

> Start a chat, attach files, reference your content, turn on deep research, answer the AI's questions and approvals, and work with its answers.

Every piece of work in Weelkit starts in a chat. The AI can answer questions, research, and create or edit articles and other writing. Articles and writings it creates open in the editor next to the chat.

## Start a chat

1. In the sidebar, choose **New chat** (Designers choose **Weelkit AI**). You can also choose **New Chat** on the Dashboard.
2. Describe what you need in the message box. Be specific about the audience, length, angle and sources you expect.
3. Press **Enter** to send. Use **Shift** + **Enter** for a new line.

While the AI works, the chat shows how long it has been working, for example "Working for 9s". Expand it to see each step, with a short title for what the AI is doing. For bigger requests, a **Step 2 / 3** label above the message box shows its plan; point at it to see the steps.

To stop, choose the square stop button. The chat says **Stopped.**, and what was already done, if anything.

## Add context to your message

> **Attach files or photos**
>
> Choose **+** (**Add options**), then **Add files or photos**. You can attach up to 10 items per message, each up to 10 MB: PDF, DOCX, XLSX, PPTX, CSV, TXT, Markdown, JSON, PNG, JPG and WEBP. Each file shows a chip while it uploads, then "Finishing upload…". If an upload fails, retry or remove it; if you send anyway, the message goes without it.

> **Reference your articles, knowledge and images**
>
> Type **@** and choose an item from your **Library** (articles and writings), **Knowledge** or images. Use the arrow keys and **Enter** to pick, or **Esc** to close the list. The AI reads what you reference. You only see items you have access to, and references count toward the 10 items per message.

> **Choose an editorial category**
>
> Choose **+**, then **Select category** to apply one of your organization's editorial category rules to the work. The category shows as a chip above the message box. This option is for Managing Editors and Admins, and the categories come from [Organization settings → Editorial](https://docs.weelkit.com/settings/organization.md#editorial).

> **Turn on deep research**
>
> Choose **+**, then turn on **Deep research**. The AI runs many searches, reads the sources, and writes a cited report, saved as a writing. It takes longer than a normal answer. A **Deep research** card shows its progress; choose the card, or **View sources** in its menu, to see its sources and searches.

> **Dictate**
>
> Choose **Start voice input** and speak, for up to 2 minutes. Then choose **Stop recording and transcribe**. The text appears in the message box for you to review before sending.

## When the AI needs something from you

* **A question:** the AI may ask a short question with options. Choose an option (or press its number), or type your own under **Something else**, then choose **Send answer**. Choose **Skip** to let the AI continue with a sensible default.
* **An approval:** before anything leaves Weelkit, such as sending an email or a Slack message, or publishing to your CMS, the AI shows what it will send and where. Nothing is sent until you choose **Approve**; choose **Deny** to refuse. For CMS sends and external tools you can choose, under **Remember this approval**, to let the AI repeat that action without asking **For 15 minutes**, **For 1 hour** or **For the rest of today**, or keep **Just this once**.

If you close the approval without deciding, an **Approval needed** label reopens it. While approvals are remembered, a label shows how many; choose it to **Stop** one or **Stop all**. You can also manage them in [Account settings → AI Preferences](https://docs.weelkit.com/settings/account.md#ai-preferences).

Gmail and Slack sends always need approval for the exact message and recipient or channel. **Remember this approval** is not offered for those sends; changing the message or destination needs a new approval.

## Work with answers

* **Copy** copies a message.
* **Regenerate**, on the latest exchange, asks the AI to answer again. When an answer has several versions, use **Previous version** and **Next version** to switch between them.
* **Edit** is on your latest message. Edit the text and choose **Send**: Weelkit sends it as a new version of that message, and the earlier one stays viewable.

Point at your own message to see its actions. Choose **Load earlier messages** to see the start of a long chat.

The number of sources an answer used shows under it; choose it to see them. To see every source of the chat, open **Sources**: **Sources** lists the web pages and knowledge the AI read, **Cited** what it cited, and **Files** the files uploaded or created in the chat.

> [!NOTE]
>
> AI can make mistakes. Check facts, figures and quotes before you publish.

## The chat and the editor

When the AI creates or edits an article or writing, a card for it appears in the chat. The document opens in the editor next to the chat; choose **Open** on the card to open it again. Use **Close Editor** and **Open Editor** to hide or show the editor.

You can also open any article you can read in your own chat, from the [Library](https://docs.weelkit.com/library-and-knowledge/library.md). If you can only read or comment on it, the AI can read it and answer questions, but won't change it.

## If something goes wrong

* If a chat was deleted or belongs to another account or organization, the page says **This chat isn't available**. Choose **Open Dashboard** to find your other chats, or **New chat** to start again. If the chat couldn't be loaded, check your connection and choose **Try again**.
* If a message fails, the chat says why and offers **Try again** when retrying can help. If deep research isn't available, you can choose **Send without deep research**.
* If a run is interrupted, the chat records what happened and whether anything was changed. Later messages don't pick the interrupted work up again unless you ask.
* If AI is turned off for your organization or your role, or a monthly AI limit has been reached, the chat says so and isn't sent to the AI. An admin can change this in [Organization settings → AI & Agents](https://docs.weelkit.com/settings/organization.md#ai-and-agents). A reached limit resets on the date the message gives.

## Find earlier chats

Your chats are listed on the **Dashboard**, most recently active first. Search them by title or message, or press <kbd>⌘</kbd> <kbd>K</kbd> / <kbd>Ctrl</kbd> <kbd>K</kbd> to search everything. To rename or delete a chat, use its menu on the Dashboard. See [Find your way around](https://docs.weelkit.com/get-started/find-your-way-around.md#dashboard).


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