Frequently Asked Questions
Everything you need to know about the WIN System
What is the WIN System?
The WIN System is a native desktop AI assistant for Windows that transcribes system audio and microphone input in real-time, captures screenshots for visual context, and uses large language models to surface What's Important Now. It runs as a lightweight desktop application — not in a browser — so it can hear everything playing on your computer and see your full screen.
How much does the WIN System cost?
$29/month or $290/year (2 months free). Each subscription includes one device seat. You can add additional seats during checkout for multi-device use. See the Subscribe page for details.
Is there a free trial?
Yes! New users receive 30 free AI calls with no account or credit card required. After that, you can start a 30-day free trial by creating an account, or bring your own Groq API key to use the app indefinitely without a subscription.
What operating systems are supported?
The WIN System currently supports Windows 10 and later. It requires approximately 4 GB of RAM and 200 MB of disk space. A ~150 MB AI model is downloaded on first launch for local audio transcription. macOS support is planned for a future release.
Does the WIN System record my audio and send it to the cloud?
No. Audio transcription is performed entirely on your device using a local Whisper AI model. Your audio never leaves your computer. Transcripts and screenshots are only sent to Groq (the AI provider) when you explicitly click "Ask the AI." You can also use your own API key for full control over where your data goes.
Can the WIN System modify files or take actions on my computer?
No. The WIN System is strictly read-only by design. It observes (audio, screenshots) and analyzes, but it cannot click buttons, fill forms, modify files, send emails, or take any autonomous actions. This is an intentional safety feature — the human always stays in the loop.
How is this different from browser-based AI agents?
Browser agents (like OpenClaw systems and ClawdBots) run inside a web browser and can only interact with web pages. The WIN System runs at the operating-system level, giving it access to system audio, microphone input, and full-desktop screenshots — capabilities no browser agent can replicate. Read our detailed comparison for more.
What is RAG and how does it work?
RAG (Retrieval-Augmented Generation) lets you upload documents (PDF, TXT, Markdown, CSV, etc.) to the WIN System. When you ask the AI a question, it automatically searches your documents for relevant snippets and includes them in the AI prompt. This gives the AI access to your private knowledge base — product docs, pricing sheets, internal guides — for more accurate, contextual answers.
Does the WIN System integrate with CRMs?
Yes. The WIN System has built-in integrations with Salesforce and HubSpot. It can structure AI responses as CRM-ready records — Leads, Contacts, Cases, Tasks, Deals, Opportunities, Events, and more. It also supports CRM Read Mode for querying existing records. Zoho and Dynamics 365 support is coming soon.
What AI model does the WIN System use?
AI analysis uses Llama 4 Scout (17B) hosted on Groq for fast inference. Audio transcription uses an on-device Whisper model that runs locally — no cloud dependency. You can provide your own Groq API key for direct access, or use the built-in proxy included with your subscription.
Does the WIN System have an API?
Yes. When the desktop app is running, it exposes a local HTTP bridge on localhost:5010. You
can start/stop recording, trigger AI analysis, retrieve transcripts, and clear the buffer — all
programmatically. This enables integration with scripts, automation tools, and custom workflows. See the API Docs for endpoints.