Frequently asked.

AssistantX (codename: Jarvis) is a personal AI assistant that sits on your machine — not a chatbot in a browser tab. It listens to voice, reads context off your screen, controls your desktop, remembers what you tell it, and gets better the more you use it. Think of a real assistant that happens to live in software.

Yes. AssistantX is local-first: the assistant, its memory, and its skills run on your machine using local models (Ollama and similar). No internet, no problem — your data doesn't leave your computer, and the assistant keeps working. When a task truly benefits from a cloud model, it can escalate to one, but only with your consent.

By default it uses local open-source models. When a task needs more horsepower, you can enable escalation to cloud providers (OpenAI, Anthropic, Google, OpenRouter). Every task is automatically routed to the best-fit model for the job, and you can pin favorites or force local-only if you prefer strict privacy.

Yes — this is one of the core ideas. Every failure and success is analyzed, classified, and turned into a lesson. Skills version like software: they improve with use, can be rolled back if a new lesson misfires, and every learned change is inspectable. There's also anti-poisoning validation so the assistant can't be tricked into learning something harmful.

AssistantX has layered memory — conversation memory for the current thread, and long-term memory for facts, preferences, projects, and decisions across sessions. Every memory is inspectable, editable, and deletable. Nothing is a black box: you can open the audit trail and see exactly what the assistant remembers and why.

Yes. AssistantX operates the desktop natively — launching apps, clicking, typing, moving files, coordinating workflows across your tools. Every real-world action passes through a permission gate you control, so nothing happens behind your back. You can also disable desktop control entirely and use AssistantX purely as a voice-chat assistant.

By design. Local-first architecture means your workspace, memory, files, and prompts stay on your machine unless you explicitly send them to a cloud model. Your private content is never used to train foundation models. Data in transit is encrypted, storage is authenticated, and you can wipe your workspace or account at any time — details in the Privacy Policy.

Those are AI models. AssistantX is the assistant around them: persistent memory, real desktop actions, offline operation via local models, learning from experience, voice-first interaction, and multi-agent reasoning for complex tasks. You can use those models through AssistantX — but AssistantX brings the parts they don't: memory, autonomy, execution, and continuity.

For big tasks, AssistantX can spin up specialist sub-agents that work in parallel — a researcher, a coder, a reviewer — coordinated by a trust-scored orchestration layer with adversarial verification of results. Instead of one model guessing alone, several agents cross-check each other. Arriving in v2.0.

Desktop app for Windows, macOS, and Linux (packaged with Electron). A web workspace runs in any modern browser. A native mobile app is on the roadmap.

We're onboarding waitlist members in waves — join the waitlist above to get an early invite. A generous free tier will always exist for everyday use. Paid plans unlock heavier cloud-model usage, premium model access, and priority queues. Pricing is always shown inside the app before you commit.

Use the Contact page (also linked in the footer) — pick a topic and your mail app will open with the right subject pre-filled. We aim to reply within 1–2 business days. You can also join the Discord community for real-time discussion.