Download the app
Install Slynkz — the Mac build is available today — and create your account. Free to try with your own local models (Ollama) — no card required. Managed cloud models start on paid plans.
Compound intelligence
A private AI workspace that lives on your own computer. It remembers what matters, builds the tools you need, and acts only with your permission — sealed in a box nobody else can open. Including us.
Free to try with local models · No card required · macOS today, Windows and Linux in development · v0.1.32
Memory
Not a chat log — distilled understanding. Slynkz keeps what it learns as durable facts that ride into your next conversation, so it gets more useful the longer you use it.
Capabilities
Say what you want in a sentence and it writes the tool, then runs it in a sandbox on your own machine. Point it at something you already have, or hand it your own code — the way in is the same.
Permission
An assistant that can use tools is only worth having if you can see what it did and stop it. Both are built in, and neither is a setting you have to go and find.
The first time the assistant wants to use a tool, it stops and asks. Nothing runs unattended.
Approve it or refuse it. Approve the same tool ten times in a row and it offers to stop asking — your choice, not a default.
Every run shows up in the chat it belongs to and in the activity log, including the ones you already allowed.
One list holds everything you've allowed. A single switch stops every tool at once, and it stays stopped through a restart.
Earned trust never extends to workflows or schedules — those have their own gates.
The box
A workspace is a sealed container on your own machine, and the AI works inside it. Nobody outside it can read what is in there — including us. Not as a policy we could revise later, but because we do not hold the key.
Your workspace is encrypted at rest and the key lives in your operating system's keychain. It works with no network at all.
The cloud holds sealed blobs and key references — never anything it can read.
Private from other users, but readable by our cloud. That is the honest description, and it is why this rung is named the way it is.
The same sharing, end-to-end encrypted so we cannot read it either. It needs key material that does not exist in the system yet.
Two workspaces agreeing to exchange sealed packages — a branch and a head office, or a vessel and its base.
The whole thing running on hardware you own, in premises you control. There is no self-host build today — no server packaging and no operator documentation.
Every rung is labelled for what it actually does today, including the one where our cloud can read. We would rather name that plainly than describe a level of privacy the architecture does not deliver yet. Key material stays bounded the same way: platform keys never reach the app, and your own secrets stay behind the Keychain and the capability boundary.
Read the privacy modelHow it works
Install Slynkz — the Mac build is available today — and create your account. Free to try with your own local models (Ollama) — no card required. Managed cloud models start on paid plans.
Use managed credits, local models, or approved capability connectors for providers and tools.
Chat, build capabilities, and run your workspace — privately, on your own machine.
Pricing
Every plan includes the local workspace and account portal. Paid plans add the managed model gateway with monthly credits and top-ups.
€0/mo
No managed credits
Get started€29.99/mo
250,000 managed credits
Get started€99.99/mo
1,000,000 managed credits
Get startedFAQ
Yes. Your workspace lives in an encrypted database on your own computer, with the key held in your OS keychain. Cloud sync, when enabled, stores only ciphertext — Slynkz cannot read your content.
OpenAI, Anthropic, Google, Mistral, Groq and more through managed routes, local models via Ollama, Brave for web search, and MCP-compatible tools through approved capability connectors.
No. Paid plans include managed model credits through the Slynkz gateway. Local models and connector secrets are configured through local or capability boundaries, not raw key fields in the web UI.
Credits meter managed model usage. Plans include a monthly allowance, and you can buy top-ups anytime from the billing portal. Local models you run yourself don't consume credits.
Your local workspace and on-device models work without a connection. Managed models and sync need the network, but your data stays local either way.
The Mac build ships today, for Apple Silicon. Windows and Linux are in development, and mobile and tablet are planned — the workspace itself is not Mac-specific, and your account and credits carry across every platform.
Install Slynkz on your desktop. Everything it learns lives on your machine, under your key, and leaves only when you say so.
Not on a Mac? See platform status — Windows and Linux are in development, with mobile and tablet planned.