AI setup

Cloud Discovery is bring-your-own-key (BYOK). AI features run on your own OpenAI or Anthropic API key, and you pay the provider directly for what you use. This page covers getting a key, choosing a model, and what to expect in costs.

How BYOK works

There is no AI subscription in Cloud Discovery and no middleman. When you use an AI feature, the app sends the request directly from your Mac to your chosen provider, authenticated with your key. That means:

• Your key, your billing. Usage appears on your OpenAI or Anthropic account at provider rates, with no markup.
• Your data goes to your provider. The relevant project context is sent from your machine to the provider's API under their data terms. It never passes through any ZoniteAI server, because there are none.
• AI is optional. Most of the pipeline is deterministic and works without a key at all. See what works without an AI key.

Get an API key

OpenAI

1. Sign in at platform.openai.com/api-keys.
2. Add billing credit to your account if you have none. API usage is prepaid or pay-as-you-go, separate from ChatGPT subscriptions.
3. Create a new secret key and copy it immediately. OpenAI shows it only once.

Anthropic

1. Sign in at console.anthropic.com.
2. Add credit under Plans & Billing if your account has none.
3. Create an API key in the API Keys section and copy it immediately.

Add the key to Cloud Discovery

1. Open Settings from the sidebar.
2. In AI Configuration, choose your provider: OpenAI or Anthropic (Claude).
3. Pick a model, or keep the preselected balanced default.
4. Paste your API key and click Save Settings.

The key is stored in the macOS Keychain, never in the database or a plain file, and it is only ever sent to the provider you selected. The app does not make a test call when you save: the key is first exercised when you run an AI feature. If the stored key looks like it belongs to the other provider (Anthropic keys start with sk-ant-), Settings shows a warning so you can correct it before generating.

Choose a model

The preselected balanced model is the right pick for most reports. Switch to a faster model to save cost, or a deeper reasoning model for large or complex environments.

You can change the model in two places: in Settings, or from the model picker in Scout's chat composer. Both write the same setting, so a change made in the chat applies to all AI features across the app: summary, documentation, migration analysis, and the topology AI Architect. Switching providers is done in Settings only.

Typical costs

A full documentation run typically costs a few cents on a balanced model. Costs scale with the size of your environment, the number of sections you generate, and the model you choose: fast models cost less, deep reasoning models cost more.

• The AI Usage page (Library section of the sidebar) records every AI call locally: tokens per month, per project, per operation, and per model.
• The app shows token counts, not dollar amounts. For exact billing, check your provider's dashboard: OpenAI under Billing, Anthropic under Plans & Billing.

What works without an AI key

Everything deterministic runs without a key and without an internet connection:

• File upload and parsing (CSV, JSON, Excel), plus live Azure discovery and deep scan.
• Resource classification across 120 or more Azure resource types.
• Relationship mapping and the findings engine (security and reliability observations).
• The topology diagram: auto-generated layout, editing, and PNG, PDF, and JSON export.
• Migration assessment: the full rules engine, dependency groups, complexity rating, and CSV export.
• Exporting documentation that was already generated, in any visual style.

These features require a license and an API key:

• Environment Summary.
• Documentation generation and per-section regeneration.
• Scout, the AI Analyst.
• The AI migration analysis report.
• The topology AI Architect.

Next step

Run your first assessment end to end.