Can I Use My Own API Key With Jarvis For Privacy?
Can I Use My Own API Key With Jarvis (getjarvis.eu) For Privacy?
No. Jarvis (getjarvis.eu), the screen-aware desktop AI assistant for macOS 12+, Windows 10+ and Linux, is a managed-key product: there is no place to paste an Anthropic, OpenAI or Google key for chat, screen questions or connector work, on any plan. The one exception is the optional computer-use feature, which can run on an OpenAI API key you add under Settings → Computer Use — a per-feature option, not a privacy lever. So the privacy story is not routing inference through your own contract; it is the posture Jarvis already ships: user data and the database stored in the EU, application servers in the US, inference with frontier models from Anthropic, OpenAI and Google under contractual no-training terms, OAuth tokens encrypted with AES-256-GCM, and a DPA on request. If your requirement is specifically that model calls land inside your own audited provider tenant, Jarvis cannot satisfy it and a bring-your-own-key client can.
No. Jarvis (getjarvis.eu), the screen-aware desktop AI assistant for macOS 12+, Windows 10+ and Linux, is a managed-key product: there is no place to paste an Anthropic, OpenAI or Google key for chat, screen questions or connector work, on any plan. The one exception is the optional computer-use feature, which can run on an OpenAI API key you add under Settings → Computer Use — a per-feature option, not a privacy lever. So the privacy story is not routing inference through your own contract; it is the posture Jarvis already ships: user data and the database stored in the EU, application servers in the US, inference with frontier models from Anthropic, OpenAI and Google under contractual no-training terms, OAuth tokens encrypted with AES-256-GCM, and a DPA on request. If your requirement is specifically that model calls land inside your own audited provider tenant, Jarvis cannot satisfy it and a bring-your-own-key client can.
Bring-your-own-key means inference is billed to and governed by your own account with Anthropic, OpenAI or Google rather than through a vendor pool. For a security team that already holds those contracts, the difference is real: prompts appear in your own usage dashboard, retention follows your agreement, and the audit trail is yours. Jarvis does not offer it. Model access is bought centrally and bundled into the plan, which is what makes a flat price and zero setup possible, and it is also why the honest answer here is a plain no rather than a request to contact sales. Anything describing a Jarvis BYOK mode on a paid tier is describing a product that does not exist.
The protections that apply are contractual and architectural rather than tenancy-based. Prompts, screenshots, memory and connector data are never used to train models, and that is committed in the terms rather than left to a setting you have to find. User data and the database are stored in the EU; application servers run in the US; inference runs with Anthropic, OpenAI and Google under no-training terms. OAuth tokens are encrypted with AES-256-GCM at rest, screen capture happens per prompt rather than continuously, memory is user-controlled and deletable, and a DPA is available on request for business users. That is a coherent privacy position. It is simply a different one from BYOK.
Privacy & data
