Local-first
Everything runs on your device. The assistant, the language model, the document index, the database — all of it lives and computes on hardware you control. No cloud dependency, no account requirement, works offline.
Silo — private, local-first AI
Silo builds an AI assistant and the full stack beneath it — inference, models, security, sync — engineered to run entirely on hardware you own. Not privacy-by-policy. Privacy by architecture.
This page: 0 third-party requests 0 cookies 0 JavaScript — open DevTools and check.
Everything runs on your device. The assistant, the language model, the document index, the database — all of it lives and computes on hardware you control. No cloud dependency, no account requirement, works offline.
No data leaves the device. Your conversations, files, and keys never touch our servers — we don't have servers that could hold them. When data must cross machines, it travels end-to-end encrypted with keys only you hold.
Five components, one guarantee. Each piece is useful alone; together they are a complete private AI stack.
A personal AI assistant for your documents, browser, and tools — running against a model on your own machine, never anyone else's.
A local LLM server with an OpenAI-compatible API. Your applications get modern AI; your prompts stay on your hardware.
Signed, verifiable model distribution. Cryptographically attested manifests and delta updates, so you always know exactly what weights you run.
The trust layer: hardware-backed key storage, encrypted databases, and audited crypto primitives shared by every Silo component.
Zero-knowledge sync between your devices. End-to-end encrypted with your keys — the relay sees ciphertext and nothing else.
A silo holds what you put in it. Ours does the same — the boundary is your device, and nothing inside crosses it.
One install puts the assistant, a local model, and encrypted storage on your laptop. First run downloads the model once — after that, it works offline.
Questions, documents, and answers flow between processes on your machine. Inference happens on your CPU and GPU; storage is encrypted at rest with keys in your device's secure hardware.
Add a second device and Silo Sync moves your data as ciphertext, end-to-end encrypted. No one in the middle — including us — can read it.
Questions, early access, or partnership — write to hello@silodev.net. A person reads it on a machine they own.