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

# Chat

> Start a conversation with a Project agent and return to its history.

Use **Chat** for a conversation in a Project workspace. Use an [Issue](https://docs.longoz.ai/issues.md) when the task needs acceptance criteria, status tracking, or a review workflow. Conversations keep their own messages and turns; sending a chat message does not itself create an Issue.

## Start a conversation

1. Open **Chat** and choose **New chat**.
2. Choose **Select project**, then **Select agent**. The agent picker lists runnable agents for that Project.
3. Write your request and choose **Send message**. Enter sends the message; Shift+Enter adds a line break.

If there are no runnable agents, check the [Project](https://docs.longoz.ai/projects.md), its [Agent](https://docs.longoz.ai/agents.md) assignments, and the assigned [Computer](https://docs.longoz.ai/computers.md). Changing the Project before sending clears the agent selection.

For a Project containing a repository, try:

> Explain how this repository handles sign-in. Identify the entry points and relevant files. Do not change any files.

This is a task for your configured CLI agent. It can use only the tools, files, and authenticated accounts available on its Computer. Chat requests can cause tools to run, so state explicitly when you want an explanation without edits.

You can also choose an enabled workspace conversation starter. The **Skills** control reflects the selected agent's configuration; manage the agent's [Skills](https://docs.longoz.ai/skills.md) there.

## Add context and follow the reply

Use **Attach files** or drop files into the composer when attachment permissions allow it. Wait for uploads and the security scan to finish before sending. If sending fails, Longoz keeps your draft and attachments.

Each message starts a turn. Wait for the current turn to finish before sending another message. Use **Stop** to stop that turn; existing messages remain in the conversation.

If a reply fails, inspect its output before choosing **Retry turn**. A turn can run tools even when it finishes without a reply, so retrying may repeat work.

## Return to a conversation

Use **Search conversations** to find earlier chats and open one to continue it. The conversation retains its Project and agent context. Load earlier messages when you need more history.

**Archive conversation** removes a chat from active history and prevents further messages. Its recorded messages, turns, Runs, and session history remain. Start a new chat for new work after archiving.


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