A bring-your-own-API-key AI assistant for Windows
AuraCloak runs on your own API key for OpenAI, Anthropic Claude, Google Gemini or Azure OpenAI, your GitHub Copilot sign-in, or any OpenAI-compatible endpoint such as Ollama or OpenRouter. Keys are encrypted on your PC with Windows DPAPI and prompts go straight to the provider. You pay the provider for what you use, and AuraCloak stays $9 a month or $49 once.
Who pays for the AI, and where your prompts travel
Most AI overlays bundle the model into the subscription. The vendor picks the model, buys the tokens and passes the cost on, sometimes with daily response caps on cheaper plans. A bring-your-own-API-key AI assistant flips that. You open an account with OpenAI, Anthropic or Google, paste the key, and pay the provider's normal rates for exactly what you use.
Two details matter more than the label. First, the request path. Some BYOK apps still send your requests through their own servers; Raycast's manual says it does this to unify model APIs. AuraCloak calls the provider directly from your PC, with no account, no telemetry and no chat sync on our side. Second, the contract. With your own key, your agreement with the provider governs how your data is handled, which is why companies with an Azure OpenAI deployment can point AuraCloak at their own tenant.
And if you'd rather not manage keys at all, browser mode uses your ChatGPT Plus or Claude Pro login instead.
Every way to connect a model
Model names are examples. Use any model your account can access.
| Provider | How you connect | Example models | Screenshots (vision) |
|---|---|---|---|
| OpenAI | API key (sk-...) | gpt-4o, gpt-4o-mini, o3-mini | gpt-4o and later |
| Anthropic Claude | API key (sk-ant-...) | claude-opus-4-5, claude-3-5-sonnet-latest | Claude 3 and later |
| Google Gemini | Free API key from Google AI Studio, or Google sign-in (OAuth) | gemini-2.0-flash, gemini-1.5-pro | All Gemini models |
| GitHub Copilot | Sign in with GitHub, uses your Copilot subscription | Fetched live from your plan | gpt-4o |
| Azure OpenAI | Resource name, deployment name and key | Your deployment, such as gpt-4o | gpt-4o deployments |
| Custom endpoint | Any OpenAI-compatible URL, key optional | Ollama, LM Studio, OpenRouter, vLLM | If the endpoint accepts OpenAI vision format |
| Browser mode | Your normal web login, no key | ChatGPT, Claude, Gemini, Copilot, Perplexity web | Through the site's own upload |
Which key fits which situation
Your company runs Azure OpenAI
Enter the resource and deployment names. Prompts stay inside the tenant your security team already approved.
Your employer pays for Copilot
Sign in with GitHub and use the models on that subscription for questions outside the editor too. See AI coding overlay.
Data that can't leave the machine
Run Ollama or LM Studio and point a custom endpoint at localhost. Prompts and screenshots stay on your PC.
Many models, one key
Use OpenRouter as a custom endpoint and change models by editing one field.
Cheap model for quick questions
Configure two providers and flip between them from the title bar: a fast, cheap model on calls, a stronger one for code review.
Trying Gemini for free
Google AI Studio gives out free API keys. Paste one in to try a model before you set up billing anywhere else.
A local model in three fields
With Ollama running, this is the whole custom endpoint setup. AuraCloak streams the reply exactly as it does for OpenAI.
- Any server that speaks the OpenAI
/chat/completionsstreaming format works - Leave the key blank when the server doesn't ask for one
- Your custom system prompt and the last 10 exchanges are sent as context, same as cloud providers
Add your key in four steps
Open Settings
Right-click the tray icon and choose Settings, or click the gear in the overlay.
Pick a provider
On the AI Providers tab, choose OpenAI, Claude, Gemini, Copilot, Azure or Custom.
Paste and save
Enter the key and model name, then Save. The key is encrypted with DPAPI before it touches disk.
Switch any time
Configured providers show a green dot in the title bar dropdown. The switch is instant.
Bundled AI, proxied BYOK and direct BYOK
| Bundled-credit AI overlays | BYOK apps that proxy requests | AuraCloak | |
|---|---|---|---|
| Who bills AI usage | The app vendor | Your provider | Your provider |
| Request path | Vendor's servers | Vendor's servers, then provider | Your PC straight to the provider |
| Model choice | Vendor's list | Vendor's supported list | Any model on 6 provider types |
| Company-hosted or local models | No | Varies | Azure OpenAI, Ollama, LM Studio, vLLM |
| Usage caps | Set by the vendor's plan | Your provider's limits | Your provider's limits |
| Account with the app | Required | Usually required | None, just a license key |
Use your own key in an overlay that stays out of screen shares
$9 a month or $49 once. AI usage is paid to your provider directly.
Bring your own API key questions
Where are my API keys stored?
On your PC, in %APPDATA%\AuraCloak\settings.json, encrypted with Windows DPAPI for your Windows user account. Keys are never written to auracloak.ini and never sent to AuraCloak.
Does AuraCloak see my prompts?
No. AuraCloak has no account system, no telemetry and no chat sync. Prompts go directly from your PC to the provider you picked. An optional local debug log can record prompts, but it is off by default and stays on your PC.
How much will API usage cost?
It depends on the provider, the model and how much you ask. You pay the provider at its normal rates, and AuraCloak adds nothing on top. Set a spending limit in the provider's billing console. Browser mode costs nothing extra if you already have ChatGPT Plus or Claude Pro.
Can I use a local model instead of a cloud API?
Yes. Choose Custom and enter any endpoint that accepts the OpenAI chat completions streaming format, such as Ollama at http://localhost:11434/v1/chat/completions, LM Studio or vLLM. Leave the API key blank if the server doesn't need one.
Can I use my GitHub Copilot subscription?
Yes. Choose GitHub Copilot in Settings and click Sign in with GitHub. Enter the one-time code at github.com/login/device. AuraCloak then lists the models available on your Copilot plan.