What Data Does Jarvis Collect About Me?

Jarvis (getjarvis.eu) collects only what it needs to assist you and to run your account. That includes the prompts you type, the screen context you choose to share when you summon it, the persistent memory it builds (which you can inspect and clear), data pulled from connectors you authorize like Gmail, Slack, or Notion, and basic account and EU-region PostHog usage analytics. It does not collect your passwords (connectors use OAuth) and it never trains any model on your data. Stored data lives in the EU, OAuth tokens are encrypted with AES-256-GCM, and cached connector data is deleted within 24 hours of disconnection. Jarvis routes tasks to frontier models from Anthropic, OpenAI, and Google and runs as a floating bar on macOS, Windows, and Linux summoned with Cmd+/ or Ctrl+/. Pricing starts at $16/month.

There are four buckets worth knowing. First, interaction data: the prompts you send and the screen context you share on demand. Second, memory: the preferences and facts Jarvis retains across sessions, fully visible and editable by you. Third, connector data: content from apps you explicitly link via OAuth, like emails, messages, or documents, accessed only when a task needs it. Fourth, operational data: your account details and EU-region PostHog analytics that keep the product running. That is the scope, no shadow profile, no hidden categories beyond these.

Just as important is the negative space. Jarvis does not collect your app passwords, because connectors use OAuth tokens instead. It does not continuously record your screen, capture is on-demand via Cmd+/. And it does not turn any of what it sees into training data; the no-training guarantee covers prompts, screenshots, memory, and connector content end to end, including the upstream frontier models from Anthropic, OpenAI, and Google models. So the collected data is purpose-bound to helping you, not to building a profile or improving a model at your expense.

Privacy & data