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

# Choose how to use SelfTune

> Compare the agent skill, Desktop app, CLI, Cloud, and self-hosted service.

Start locally. Choose the agent skill for natural-language operation, Desktop
for a visual local workflow, or the CLI for automation. Add Cloud or self-host
only when you need a shared remote library or team workflow.

| Option         | Choose it when                                                     | Where data lives    | Start here                                                                              |
| -------------- | ------------------------------------------------------------------ | ------------------- | --------------------------------------------------------------------------------------- |
| Agent skill    | You want your agent to operate SelfTune for you                    | Local by default    | [Set up with your agent](/setup-with-your-agent)                                        |
| Desktop app    | You want installation, status, and skill management in a native UI | Local by default    | [Download the latest release](https://github.com/selftune-dev/selftune/releases/latest) |
| CLI            | You need scripts, CI, a server, or explicit command control        | Local by default    | [CLI and headless use](/run/cli)                                                        |
| Self-host      | Your team needs a private remote library you operate               | Your infrastructure | [Self-hosting](/guides/self-hosting)                                                    |
| SelfTune Cloud | You need hosted review, distribution, or contributor workflows     | SelfTune Cloud      | [Cloud overview](/cloud/overview)                                                       |

## Agent skill

This is the recommended interactive path. Install the SelfTune skill, then ask
your agent to initialize, inspect health, or improve a named skill. The agent
uses SelfTune's operating instructions and invokes the local runtime for you.

Choose this when you prefer goals such as “show me which skills need attention”
over remembering commands.

## Desktop app

Desktop runs the local SelfTune service and dashboard together. It is the
clearest choice when you want to browse installed skills, manage local settings,
and keep background health checks visible without working in a terminal.

Installers are published for macOS, Windows, and Linux on the
[GitHub releases page](https://github.com/selftune-dev/selftune/releases/latest).

## CLI and headless use

The CLI is the operational API used by the agent skill and Desktop app. Use it
directly for CI, servers, scheduled jobs, or debugging. Start commands through
`npx selftune@latest` if you do not have a global executable.

The CLI is complete, but it should not be your first learning path. Begin with
an outcome guide, then use the command reference when you need exact flags.

## Self-hosted service

Self-host packages the dashboard and Remote Library protocol in one container.
Choose it when your organization needs shared immutable skill revisions,
private sharing, backup, and multi-device sync on infrastructure you control.
Raw transcripts do not sync to the remote library.

## SelfTune Cloud

Cloud is an optional creator and team surface. Use it when a skill needs hosted
review, distribution, shared evidence, or contributor signals beyond one local
installation. Local SelfTune remains the default for personal iteration.

<Warning>
  Connecting a remote library or Cloud changes the data boundary. Review
  [Privacy and permissions](/guides/privacy-and-permissions) before connecting
  an account or uploading any artifact.
</Warning>

Choose the smallest form that satisfies the user's goal. Prefer local-only
operation unless the user explicitly asks for shared remote state. Do not infer
permission to create an account, upload data, enable billing, expose a service,
or configure credentials. Before recommending Desktop, confirm the operating
system and architecture. Before recommending self-host, confirm the user is
prepared to operate TLS, authentication, backup, and upgrades.