Models and API keys
Ploof Studio writes your site with an AI model. It runs on your Setapp AI credits from the start, and you can add your own API key to run on your own account instead. This chapter covers both, and how to pick the right model for a project.
Ploof Studio writes your site with your Setapp AI credits. There is nothing to connect. Each reply spends credits, and they renew every period.
The Setapp AI row at the bottom of the sidebar shows how many credits are left, with a bar under the number.
- While you have plenty, the row stays neutral.
- Under 10 percent, it turns amber and shows a warning icon.
- At zero, it turns red.
Select the row to open Settings on Providers. The Setapp AI card there shows how many credits are left this period and when they renew, with a Get more credits link.
The card can also show one of these states.
| State | What to do |
|---|---|
| Out of credits | Your credits are used up for this period. Select Get more credits, wait for them to renew, or add an API key. |
| Not signed in | Open Setapp and sign in, then select Check again. |
| Offline | Check your connection, then select Check again. |
| Not working | Select Check again, or add an API key below the card. |
An API key is optional. Add one if you want replies billed to your own account with Anthropic, OpenAI, Vercel AI Gateway, or OpenRouter instead of your Setapp AI credits. One key is enough.
| Provider | Models | Key starts with | Get a key |
|---|---|---|---|
| Anthropic | Claude models, billed by Anthropic | sk-ant- | Anthropic keys |
| OpenAI | GPT models, billed by OpenAI | sk- | OpenAI keys |
| Vercel AI Gateway | Models from many labs, billed by Vercel | vck_ | Vercel AI Gateway keys |
| OpenRouter | Models from many labs, billed by OpenRouter | sk-or- | OpenRouter keys |
Create a key with the provider
Open the provider’s key page from the table above, create a key, and copy it.
Open Settings
Press ⌘ , or select Settings at the bottom of the sidebar, then select Providers.
Add the key
Under API keys, find the provider and select Add key. Paste the key into the API key field.
Save it
Select Save key. Ploof Studio checks the key with the provider before saving it, and the button reads Checking… meanwhile. Once it’s saved, the row says Connected and shows the last four characters of the key.
When Ploof Studio can’t save a key, it says why under the field. Nothing is stored until the provider accepts the key.
| What it says | What to do |
|---|---|
| That doesn’t look like an API key. | Paste the whole key, with no spaces before or after it. |
| The provider didn’t accept this key. | Check that you copied all of it, or create a new key and paste that. |
| The provider is limiting requests right now. | Wait a minute, then select Save key again. |
| Ploof Studio couldn’t reach the provider. | Check your connection, then try again. |
| Ploof Studio couldn’t open your Mac’s keychain. | Unlock your Mac, then try again. The key isn’t stored until the keychain opens. |
To use a different key, select Replace on the provider’s row, paste the new key, and select Save key.
To remove a key, select the trash button on its row, then Remove key. Ploof Studio deletes it from this Mac and stops using that provider until you add a key again. The key keeps working everywhere else, so revoke it on the provider’s site if you no longer need it.
The model picker sits in the prompt bar. It lists the models of every connected provider, grouped by provider, and ends with Manage providers, which opens Settings. If nothing is connected, it reads Connect a model.
Each model carries a tag that says how it trades speed for care.
| Tag | What it means | Models |
|---|---|---|
| Fast | Answers straight away and costs the least. Good for small edits and quick tries. | Haiku 4.5, GPT-6 Luna |
| Balanced | Answers straight away with more care. A good default for building and most changes. | Sonnet 5.5 or Sonnet 5, GPT-6.1 Sol or GPT-6 Sol |
| Deep | Reasons before it answers, so it takes longer and costs the most. Good for a first build you care about or a tricky change. | Opus 5.5, GPT-6 Astra |
While a Deep model reasons, the reply shows Thinking… with a timer. When the answer starts, that folds into a line such as Thought for 1m 12s, which you can open to read the reasoning.
A new project starts on the model the home screen shows, which is the latest project’s model while its provider is connected. On Setapp AI that is Sonnet 5 at first. With your own key it’s Sonnet 5.5, or GPT-6.1 Sol for OpenAI.
A project remembers its model. To change it, open the picker in the project’s prompt bar and choose another model, from any connected provider. The whole conversation goes to the next model, so it picks up where the last one left off.
If the project’s provider is no longer connected, the picker says so and shows its models as unavailable. Pick another model, or connect that provider again in Settings.
Ploof Studio names each new project with a separate model. Change it in Settings under General, in Text generation model. It writes short text, such as the title of each new project, so a fast model keeps it quick and cheap. Haiku 4.5 is the default, or GPT-6 Luna when OpenAI is your only provider.
Keys are encrypted with your Mac’s keychain and stay on this Mac. Ploof Studio sends each key only to its own provider, and never shows a saved key again, only its last four characters.
Stuck on something?
Write to us and a person reads it and replies.