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

# Set up the CTO

> Pick a model that can steer a live turn, optionally connect Linear, and start with a planning prompt.

The CTO is usable in a minute. There is no setup wizard: the first time you open the CTO tab you pick a model, and that's setup. Linear layers in afterward, only when you need it.

## First run

A project that has never picked a model opens the CTO tab on the model picker instead of the thread. Choose a model and the thread takes over.

**1. Pick a model**

The CTO only offers models that can take a message into a turn already underway — it is interrupted constantly, by the chats it starts and by its own wake-ups, so a model that queues your message instead of hearing it cannot hold the job. If nothing is listed, sign in to Claude, Codex, or Cursor under **Settings → AI → Providers**.

**2. Tune reasoning effort (optional)**

In CTO Settings, set reasoning effort and Fast mode. The CTO reasons across the whole project, so favor a capable model over a fast one.

**3. Connect Linear (optional)**

Connect it when you want the CTO to browse and update issues from the thread. See [Linear](https://ade-app.dev/docs/integrations/linear.md).

> [!NOTE]
>
> Picking a model is the whole of first run. Reasoning effort and Linear connect later — the CTO is meant to be a daily chat surface first.

## You don't choose a personality

There are no personality presets and no work-style toggles. The CTO has one voice, owned by ADE and fixed in its doctrine:

* It leads with the state of things, then its read on it, then what it would do.
* It stays level — no exclamation marks, no "great question", no congratulating you for asking.
* When it doesn't know, it says so, says what it would check, and then checks it.
* It disagrees once, with a reason. If you decide otherwise, it does it your way and drops it.
* It never says something is done before verifying it.

The same doctrine tells the CTO that it knows the ADE application and not only your repository. When you're lost in ADE — where a setting lives, what a tab does, why a control is greyed out — ask the CTO. It reads ADE's own documentation rather than guessing, and it hands you a deeplink to the place instead of directions for finding it.

## Change the model later

**CTO → Settings → Model** switches the running thread in place. The thread, its memory, and its context all carry over — a model switch is not a reset.

## First useful prompts

Lead with planning and summarization — that's where the CTO earns its keep.

```text theme={null}
Summarize the current project state and identify the highest-risk open work.
```

```text theme={null}
Plan how to split this feature into reviewable lanes before anyone starts coding.
```

```text theme={null}
Read Linear issue ENG-142 and recommend the first lane to create.
```

> [!TIP]
>
> Ask the CTO to remember decisions as you make them — it keeps durable memory across sessions and model switches, so you don't have to restate context next time.

> [card] [CTO overview](https://ade-app.dev/docs/cto/overview.md)
>
> What the CTO is and when to reach for it.

> [card] [Linear](https://ade-app.dev/docs/integrations/linear.md)
>
> Connect Linear so the CTO can browse and update issues.


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