# Install Token Tunnel

> Run your automations on the ChatGPT, Claude, Gemini, and Grok plans you already pay for, from your own computer.

Token Tunnel lets General Input run on the AI subscriptions you already have. You install [General Input Desktop](/token-tunnel/download), sign into your providers inside the app, and those models show up everywhere in General Input, at no extra cost.

## Why bother

Most automation tools bill you per run against their own metered API keys. Token Tunnel routes the work through the plan you already pay for instead.

- **No extra cost.** Runs go against your ChatGPT, Claude, Gemini, or Grok subscription rather than metered credits.
- **Your logins stay yours.** You sign into each provider inside the app, on your own computer. General Input never sees, stores, or shares those credentials.
- **Nothing technical to configure.** No API keys to create, no settings to get right. The app keeps itself up to date.

<Callout type="tip">
  You can connect one provider or all four. Each one you add shows up as a model
  option across the product.
</Callout>

## Install the app

<Steps>
  <Step>
    Download **General Input Desktop** from the [download
    page](/token-tunnel/download). On a Mac, drag it into your Applications
    folder and open it. It is free and auto-updates.
  </Step>
  <Step>
    Click **Sign in to General Input**. That links the desktop app to your
    workspace.
  </Step>
  <Step>
    Open **Token Tunnel** in the left sidebar of the desktop app, then sign in
    to **ChatGPT**, **Claude**, **Gemini**, or **Grok** the same way you would
    sign into a website. Each account and its tier appears once connected.
  </Step>
</Steps>

## Use your models

Once a provider is connected, its models appear in the model picker across General Input, alongside the hosted options. Pick one in any chat, workflow, or app and it runs through your own computer.

There is nothing to switch on. If you would rather a particular automation stay on hosted models, just pick a hosted model for it.

## When your computer is off

Your machine has to be awake for these models to answer. If it is asleep or offline, General Input tells you so rather than quietly falling back to a metered cloud model and charging you for it.

Turn the computer back on and the models are available again. Nothing needs reconnecting.

<Callout type="tip">
  For automations that run on a schedule overnight, pick a hosted model so a
  sleeping laptop never stalls the run.
</Callout>

## What's next

With a model in place, the next step is teaching your agents a repeatable procedure. Head to [Build a skill](/getting-started/create-a-skill).