Software shouldn't spy on the people using it.
Portico was started by engineers tired of the trade-off: you can have a capable model, or you can keep your data. We built a third option. A desktop app that runs capable AI models on your own computer, keeping everything you do firmly on your own machine.
If your AI assistant requires an internet connection to think, it isn't your assistant—it's a wiretap that types back.
The cloud is a black box.
When you type a prompt into a web UI, where does it go? Who logs it? Who trains on it tomorrow? The answer is rarely comforting for a business dealing with proprietary code, client data, or internal strategy. We realized that for AI to be truly useful at work, it had to be trustworthy first.
Built on a solid foundation.
Privacy by Default
Networking toggles are off out of the box. We don't want your data, because we never see it. The safest data is the data that never leaves the room.
Hardware You Own
Good hardware is expensive. Once you own it, you should be able to run capable AI on it without paying a monthly subscription for the privilege.
Open Source Always
Trust requires verification. Portico's core engine is open for anyone to audit. No hidden telemetry, no secret backdoors, just transparent code running on your machine.
Why we named it Portico.
A portico is a structure leading to a building, supported by columns. We thought it was a fitting metaphor. The open-weight models (Llama, Mistral, Qwen) are the columns—they do the heavy lifting. Portico is simply the elegant structure that brings them onto your computer, organizing them into tiers (Doric, Ionic, Corinthian) so anyone can step through with ease.
Try it on your own machine.
Free and open source. Works with the open models you already trust.
On a Mac or Linux? Build it in three commands →