Agents in Unstoppable Code run on the model you choose. There are three ways to run models:
- Your own subscription. Use your Claude, ChatGPT, or Cursor subscription by signing in to Claude, Codex, or Cursor.
- AI credits. Run hosted models with credits from your Unstoppable Code account.
- Local models. Run models on your own computer with Ollama.
Before you start
- Desktop app only. Ollama runs models on your own computer, so you set it up in the desktop app.
- To use AI credits, sign in to your account. AI credits belong to your Unstoppable Code account.
Sign in to Claude or Codex
- Open Settings and go to Accounts.
- In the LLM provider section, find the Claude or Codex row.
- Select Sign in.

For Claude, the app opens a Sign in to Claude dialog. Complete the sign-in flow in the terminal it shows.
For Codex, the app opens a Sign in to Codex dialog and your browser. Finish signing in there. If the browser does not open, select Open browser. To sign in with a code instead, select Use a device code. If Codex work is running, the dialog first asks to pause it: select Pause work and sign in, and check Use a device code first if you want a code.
When you're signed in, the row shows your account and plan. To switch accounts, select Change account.
Connect Cursor
- Open Settings and go to Accounts.
- In the LLM provider section, find the Cursor row and select Connect Cursor.
- The app downloads its own tested copy of Cursor. Your existing Cursor installation keeps its own updates.
- Finish signing in in your browser. If you're already signed in to Cursor on this computer, the app uses that sign-in.
When you're connected, the row shows the account you're signed in as. Your Cursor models appear under Cursor in the model picker. They run on your Cursor plan and its usage limits, not on AI credits. To use a different Cursor account, select Change account.
Grok is merged with Cursor. If you only have a Grok account, select Only have a Grok account? under the Cursor row, then Unify accounts. Merge the accounts in your browser, then connect Cursor.
Run models on AI credits
Hosted models run on AI credits, so they don't need a subscription of your own. In the model picker, a model that runs on credits shows AI Credits on its row. If you aren't signed in, a model that needs an account shows Sign in instead. Claude and Codex models can also run on credits when you aren't signed in to that provider.
To get credits and see what happens when a provider limit runs out, see AI credits and usage limits.
Set up Ollama for local models
- Open Settings and go to Ollama.
- Under Mode, choose how Ollama runs:
- Managed by Unstoppable Code starts Ollama when a local model is needed, and stops only the server process the app started.
- Manual server leaves Ollama to you. The app connects to your local server and never starts or stops it.
- If you chose Managed by Unstoppable Code and the app shows Ollama isn't installed, select Install Ollama. If Ollama is already on your computer, select Detect again under Existing installation.
- If you chose Manual server, select Download Ollama to install it yourself, then start the server.
- Under Server, select Test connection to confirm the app can reach Ollama.
Add a local model
Under Models, Installed models lists what you already have. To add one:
- In the Pull model section, select Ollama model list and find the model you want.
- Enter its exact name under Model name, such as
llama3:8b. - Select Pull.

Next steps
- Choose an approval mode to control what agents can do without asking.
- Start your first task with the model you connected.
