Unpack and install
Choose a Windows or macOS package, then follow a guided setup on your own hardware.
We're building a personal AI assistant you can set up on your Windows PC or Mac, then reach from your iPhone or Android through a familiar messenger like Telegram.
Not another empty chat box. Explore how a local-first assistant could combine the notes, messages, calendar, and preferences you explicitly allow — then ask before changing anything.
Try a sample question or type your own. Search results will show the exact fictional sources used.
This is a clickable product story, not a real AI endpoint. All names and records are fictional. Your browser sends no demo questions to our server.
Ask about a future pilotThe simple onboarding flow we are building toward — without a complicated server setup for every user.
Choose a Windows or macOS package, then follow a guided setup on your own hardware.
Connect approved notes, preferences, and a local model. Keep meaningful actions under your control.
Use a familiar messenger on Android or iPhone to access the assistant running on your host device.
A Windows PC or Mac acts as the assistant's home. Your phone can reach it through Telegram — you don't need a separate native AI application on every device.
Choose your host computer to explore the guided setup we are building toward. This preview runs only in your browser — no installer, account, or backend is involved.
Concept only · An installation package is not yet available to download.
Imagine asking your assistant to find a note, summarize a document, or help plan a task — without opening a complex dashboard. Choose an example to preview the idea.
Your assistant should reflect the documents you trust, the preferences you choose to keep, and the workflows you actually use.
A roadmap toward persistent, user-controlled context that makes assistance more relevant over time.
Context under your controlBring approved notes and documents into grounded AI workflows with relevant references.
Your files, your permissionsAutomation with boundaries: meaningful or sensitive actions should require your approval.
Local-first architectureLocal-first should come with clear boundaries. Explore the proposed paths for local inference and optional cloud reasoning.
Local inference stays on the host for processing. Messenger messages still pass through the messenger provider when remote chat is enabled.
Architecture illustration, not a security certification. External services may handle data when you opt into messenger or cloud integrations. Actual controls will be validated during product development.
MKC OnePack AI is an independent, self-funded side project, separate from the founder's employment. No packaged installer is for sale or public download yet.
Ask about early evaluationA clear picture of the product we are building and what is available today.
Ask a different questionNot yet. The technical prototype exists, but the customer installer and guided setup are under development. There is no public download or subscription.
The intended host is your Windows PC or Mac. Android and iPhone use a messenger interface to reach the assistant while the host is running and securely accessible.
Not entirely. Messages exchanged through a messenger are handled by that provider. Standard Telegram bot chats are not end-to-end encrypted. We aim to make external-data paths and permissions explicit.
The target architecture is local-first, with optional cloud reasoning if enabled by the user. Exact hardware requirements, model support and pricing are not finalized.
We welcome product feedback, technical conversations, and future pilot inquiries. MKC OnePack AI is currently a prototype-stage project.
Contact the project