> ## Documentation Index > Fetch the complete documentation index at: https://docs.vybe.build/llms.txt > Use this file to discover all available pages before exploring further. # Chat with a coworker > Use a coworker's Chats section in the Vybe web app: channels, threads, attachments, questions and models Every AI coworker has a **Chats** section in the Vybe web app, laid out like Slack: channels, with a thread for each conversation. Use it when you want to attach files, review what the coworker built, or work through a longer task at your desk. Most teams talk to their coworkers in Slack, in direct messages and channels. See [Slack](https://docs.vybe.build/channels/slack.md). > [!WARNING] > > Chats are shared. Everyone who can access a coworker can see all of its channels and threads, including the ones you started. Don't post anything there you wouldn't share with those people. To see who has access, open **Overview → Sharing**. ## Open a coworker's chats In the sidebar, expand a coworker and click **Chats**. You can also press `⌘ + K` (`Ctrl + K` on Windows) and choose **Chat with** *name*. ## Channels Every coworker starts with a **#general** channel, which can't be renamed or deleted. Add more channels to keep different kinds of work apart, such as #reports and #research. * **Switch channel:** click the channel name at the top of the chat to see the list of channels. * **Create a channel:** in that list, click **Create a channel** and type a name. Names use lowercase letters, numbers and hyphens, up to 50 characters, and must be unique. Vybe shows the final name before you click **Create**: "Channel will be created as #slug". * **Rename or delete a channel:** hover it in the list, click **…**, then **Rename** or **Delete**. Deleting a channel permanently deletes all its messages. ## Threads Each message you send from a channel's composer starts a new thread in that channel. The channel feed shows each thread's first message, how many replies it has, and when the last reply came in. * Click a thread, or its **Reply** button, to open it in the **Thread** panel and continue that conversation. * Close the panel with **Close thread** or `Esc`. * Click **Load earlier threads** to see older ones. > [!TIP] > > Start a new thread for each new task, and keep follow-ups in the same thread. Each thread is its own conversation. Anything the coworker saves to its notes is available in every thread. ## The composer Type in the box ("Ask me anything...") and press `Enter` to send. Press `Shift + Enter` for a new line. * **+** opens **Connect integration**, **Upload image** and **Upload document**. You can also drag files onto the composer or paste them. * **@** lets you point the coworker at something specific: an integration, a database, a data query, a custom API, an app, or another coworker. * The model pill (tooltip "Switch AI model") sets the coworker's model. See [Choose a model](#choose-a-model). ### Attachments | Type | Formats | Limits | | - | - | - | | Images | Common image formats. HEIC photos are converted to JPEG. | Up to 10 per message, 5 MB each | | Documents | .pdf, .docx, .xlsx, .pptx, .txt, .csv, .html, .md, .rtf, .zip | Keep each file under 50 MB and the total under 100 MB per message, so the coworker can open them | Attached files are copied into the coworker's own workspace, so it can read them, run scripts on them, and convert them. ### While your coworker is working * **Queue a message.** If you type and send while the coworker is still replying, your message waits in a queue (**Queue message**) and goes out when the coworker is ready. You can edit or remove queued messages before then. * **Stop.** With an empty composer, the send button becomes **Stop**. A stopped reply is marked "You stopped this run before it finished." ## What you see in a reply * **Steps.** Each action the coworker takes shows as one line with an icon, such as "Searched the web", "Ran command" or "Sent Slack message". Repeated actions are grouped. Click **View** on a step to see what went in and what came out. If the model reasoned before answering, **Show thinking** reveals it. * **Questions.** When the coworker needs a decision, it asks in a card above the composer: pick one option, pick several, rate on a scale, or type an answer. Many questions also offer **Something else** for your own answer. Use **Back**, **Skip** or **Skip all**, then **Continue**, or **Send** on the last question. You can press 1–9 to pick an option. The composer stays available: if you send a normal message instead, the card expires. * **Cards.** Some results arrive as cards. An artifact the coworker publishes opens in a side panel (see [Artifacts](https://docs.vybe.build/coworkers/artifacts.md)). An app it builds gets a card that links to the app. When it needs a tool, a **Connect** *name* card lets you connect an account, or **Grant** it access to one you already have. * **App previews.** When the coworker works on one of its apps, it can open the app's live preview next to the thread. On a phone, tap **Open preview**, then **Back to chat** to return. If the preview crashes after you've sent a message in the thread, Vybe reports the error to the coworker ("The preview crashed. Asking the agent to repair the app…") so it can fix the app, up to three times until your next message. **Stop** turns these automatic fixes off until you send another message. * **Feedback and cost.** Use the thumbs up or down (**Give feedback**) to rate a reply. Hover a reply and open **…** to see the credits it used. ## Choose a model New coworkers use **Auto**, which picks a model suited to each conversation. To choose yourself, click the model pill in the composer: * **Auto**: "Automatically pick the best model for the task." * **Cheap**, **Balanced** or **Smart**: a tier, shown with the model it currently uses. * **Other models**: any specific model, sortable by provider, cost, intelligence or speed. * **Effort**: how long the model thinks, from **Low** to **Max**. Higher effort uses more credits. > [!WARNING] > > The model belongs to the coworker, not to you or the thread. Changing it here changes it for this coworker for everyone. It's the same setting as **Automatic model selection** on **Overview → Capabilities**. Scheduled tasks and triggers can have their own model. For the full list of models, see [AI models](https://docs.vybe.build/reference/ai-models.md). Owners and Admins can tune Auto and bring their own model gateway in [Models and gateway](https://docs.vybe.build/organization/models-and-gateway.md). ## If your coworker can't reply > **Out of credits** > > Your organization has used its credits. Upgrade, buy more credits, or wait for the monthly refill. See [Plans and credits](https://docs.vybe.build/getting-started/plans-and-credits.md). > **Agent paused: usage limit reached** > > This coworker hit its monthly usage limit. Click **Manage limit** to raise it, or wait for the next monthly reset. See [Usage and activity](https://docs.vybe.build/coworkers/usage-and-activity.md). > **This thread is busy** > > The coworker is still working on a reply in this thread. Wait for it to finish, then send your message again. > **Message too large** > > Send the message again with fewer or smaller attachments. > **The conversation is too long** > > The thread has grown too long for the model. Start a new thread and summarize what the coworker needs to know. This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.