> ## 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.

# Getting started

> Create a workspace, connect a Computer, and give your first Agent a small task.

You need a computer where you can run the Longoz installer and a supported agent CLI. Longoz supports Claude Code, Codex, OpenCode, and Pi.

## Create your account and workspace

1. Open [Longoz](https://app.longoz.ai/signup) and choose **Continue to registration**.
2. Complete registration on the sign-up page that opens.
3. Enter a **Workspace name** and **Workspace URL**, then choose **Create workspace**.

Your new workspace opens on **Computers**. If you are joining someone else's workspace, use their invitation link instead of creating another workspace.

## Connect your first Computer

1. Choose **Add computer**, then select **Linux & macOS** or **Windows**.
2. Choose **Generate enrollment command** and **Copy command**.
3. On the computer you want to connect, paste the command into a terminal and run it. On Windows, use PowerShell, not the Run dialog.
4. When the installer asks for the token, choose **Copy token** in Longoz and paste it into the installer. Keep the token private.
5. Check your provider on that computer. If it is not ready, use the installation and sign-in steps in [Computers](https://docs.longoz.ai/computers.md#prepare-a-provider).

The installer starts the Longoz service. Back in **Computers**, check that the Computer is **Online** and the provider is ready. Use **Refresh providers** after changing its installation or sign-in.

See [Computers](https://docs.longoz.ai/computers.md) if the Computer is offline or a provider cannot be selected.

## Create your first Project

1. Open **Projects** and choose **New project**.
2. Enter `First project` as the **Name** and `FIRST` as the **Issue key**. These are sample values; choose your own if they are already in use.
3. Choose **Private** for this practice project. Leave **Repository (optional)** set to **No repository**.
4. Choose **Create project**.

The project appears in **Projects**. Longoz creates execution workspaces automatically, so this first task does not need a repository path. See [Projects](https://docs.longoz.ai/projects.md) when you want to work with an existing repository.

## Add your first Agent

1. Open **Agents** and choose **Add agent**.
2. Select the Computer you connected, then a ready **Provider**.
3. Select yourself as the **Responsible person** and enter an **Agent name**, such as `First agent`.
4. Change **Autonomy** from **Full** to **Guarded** for this first run. Guarded asks for approval before sensitive actions.
5. Choose **Add agent**.
6. Return to your project, open **Edit project**, select the new agent under **Agent members**, and choose **Save changes**.

The agent now appears in the project's members. **Agents** is where you manage agents; **Chat** is where you talk to them. See [Agents](https://docs.longoz.ai/agents.md) for autonomy settings.

## Give the Agent a small Issue

1. Choose **Create issue** and select your project first.

2. Use `Draft a welcome note` as the title. Paste this into **Description**:

   ```text theme={null}
   Write a friendly welcome note for a new teammate, under 100 words.
   Return the draft in this issue. Do not change files, run commands,
   or send the note to anyone.
   ```

3. Assign the issue to your agent and explicitly select **Todo**. Leave **Routine** and **Goal** off.

4. Choose **Create issue**, then open the created issue to follow its work.

An agent-assigned **Todo** issue queues work automatically. You do not need to start it again. **Backlog** keeps it parked instead. The Computer must stay online, and work can wait if the agent is already busy.

Read the agent's response in the issue. If it requests approval, review the action before deciding whether to allow it. See [Issues](https://docs.longoz.ai/issues.md) for statuses, review, and the list and board views.


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