Onboarding by use case
On first launch, you pick your use case: mail / ticketing, dev assistant, internal chatbot, or general assistant. Each profile has its own instructions, documents, and tools.
Back to hub
AI Projects · whytcard.ai
The AI that truly belongs to you.
whytcard.ai runs a Swiss AI model directly on your machine. Your documents, your questions, your answers — nothing goes to the cloud.
See howApertus 8B · local · 0 cloud
What it actually does
Local AI, scoped by use case.
Apertus 8B, local
The Swiss Apertus model (EPFL / ETH), served on your machine via llama.cpp. No cloud call to answer, no external fallback.
Your documents, indexed at home
Your files are chunked and indexed locally (SQLite + sqlite-vec). The AI answers with cited sources, and search never leaves your disk.
Integration Center
Enable, per profile and with explicit permission, local capabilities: files, browser, IDE, connectors, MCP servers. Everything is off by default.
Why it's different
Sovereign by design, not as an option.
Local first
One inference runtime: llama.cpp, local. No Ollama, no cloud provider, no hidden fallback.
Zero cloud
Profiles, instructions, documents, and index stay on the machine. No hosted account, no telemetry to get started.
Scoped + permissions
Every integration is tied to a profile and declared permissions. Default: everything denied, you allow explicitly.
Real project status
What runs, what's coming.
Today
Local profiles, profile-scoped chat, document RAG, integration catalog, tested Rust backend (8 crates).
In progress
Tool execution, browser and IDE companions, MCP client and server, mail / ticketing connectors.
Model
swiss-ai/Apertus-8B-Instruct-2509Independence
Zero foreign capital.
WhytCard is built in Ecublens, Canton of Vaud. A single founder, no venture capital, no foreign investor. That is what structurally guarantees no data can ever be siphoned by a foreign legal obligation. Privacy isn't an option: it's the consequence of independence.
Write to Jérôme1024 Ecublens, Vaud, Switzerland