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

# Quickstart

> Install Vibex, connect an Agent, and complete your first session.

This guide walks you through installing the desktop, adding an Agent, opening a project directory, and having the Agent finish a small, reviewable task. For your first try, use a throwaway Git repository.

## 1. Install the desktop

Download the installer that matches your platform from [GitHub Releases](https://github.com/vibex-ai/vibex/releases/latest) and install it. For full instructions and per-platform caveats, see [Install](/docs/en/install).

<Note>
  The macOS and Windows installers are currently unsigned, so Gatekeeper or SmartScreen stops you on first launch. For how to allow the app, see [Install](/docs/en/install).
</Note>

## 2. Add an Agent

On first launch, Vibex shows the **Start something new** page. When no Agent is available it tells you **No Agent has been added. Open Config Center to add one.** — click **Add Agent** to open the Config Center.

1. Find the Agent you want in the **Agent** tab list. Built-in presets include **Claude Code**, **Codex**, **ZCode**, and **OpenCode**, plus more than a dozen ACP integrations.
2. Click the **+** on the right of an entry to add it. Once added, the entry shows an enable toggle.
3. If an Agent is managed by Vibex and not installed yet, its card shows **Install**. Click it and Vibex downloads and verifies the runtime. You cannot cancel during installation; when it finishes, the button becomes **Check for updates**.
4. If an entry shows **Managed by user**, it is not in the managed directory, and Vibex uses the same-named command-line program on your `PATH` — you need to install it yourself first.

Next, expand the **Native credentials** card on the detail page to sign in (`Sign in` / `Open sign-in terminal`), then configure the model provider and credentials in the **Model provider configuration** card.

## 3. Open a project directory

1. Click **New project** in the sidebar, or use **Project directory** on the **New chat** page.
2. In the directory browser that opens, select the repository root and confirm.
3. For your first try, use a test branch or a temporary repository — the Agent really does modify files on disk.

## 4. Complete your first session

1. At the top of Composer, choose the Agent, provider profile, model, and **Thinking depth**.
2. Type a clearly scoped task, such as "add installation steps to `README.md`, and show me the plan first".
3. Watch the plan and tool calls in the timeline. When it needs files, a terminal, or permission, Vibex shows the specific request, and you can **Approve** or **Deny** it.
4. When the Agent finishes, open the **Git** panel and check **Changes**, confirm each modification, then decide whether to commit or keep asking questions.

<Tip>
  For a first task, ask the Agent for a plan before letting it modify files. That way you can confirm the scope and risk in the timeline.
</Tip>

<Warning>
  **Stop** only stops the run. It does not roll back files already written or commands already executed. After stopping, verify the result in the Git diff and in the terminal.
</Warning>

## Understanding Composer input

You can combine several kinds of context in Composer:

| Input             | What it does                                            |
| ----------------- | ------------------------------------------------------- |
| `/`               | Pick an available Prompt or slash command.              |
| `@`               | Reference a file or directory from the current project. |
| `$`               | Insert an enabled Skill.                                |
| Attachment button | Add images and other session attachments.               |

Before sending, you can choose the Agent, provider profile, model, thinking depth, and conversation mode. Different Agents support different options; the interface shows what is available based on what the runtime detects.

## Next steps

* [Core concepts](/docs/en/concepts) — how projects, workspaces, sessions, and the authoritative runtime relate to each other.
* [Workspaces](/docs/en/workspaces) — the boundary for files and Git.
* [Config Center](/docs/en/config-center) — configuring Agents, MCP, and Skills.
* [Pair mobile](/docs/en/remote-mobile) — continue these sessions on your phone.