Arlo AI is a family of open-source projects for running AI agents on your own terms — a private agent runtime for your machine, a mobile client that talks straight to your provider, and the protocol layer that connects an agent to any frontend. No backend of ours in between. Ever.
Every one of them free, MIT-licensed, and yours to run.
A private, local-first agent that speaks whichever interface you need — an interactive TUI, a one-shot CLI, or an embeddable library — and runs on macOS, Windows, and Linux. Your keys, your machine.
A React Native iOS app for talking to LLMs on the go. Bring your own API key and connect direct to your provider — no account, no subscription, nothing sitting in between.
An up-to-date Rust implementation of the AG-UI protocol — the standard, event-based interface that lets a web or app frontend drive Arlo Rust, wire-compatible with the TypeScript and Python SDKs.
Two independent paths to the same place — your own model, without a middleman.
The same four commitments across every project.
Your credentials stay on your machine and go straight to your provider. We never see them, because there is nowhere for them to go.
There is no Arlo server to sign up for, no seat to pay for, and no account to create. You pay your model provider and nobody else.
Anthropic, any OpenAI-compatible API, or a model server running on your own hardware. Point it wherever you like and switch whenever you like.
Every project is developed in public under the MIT license. Read it, fork it, ship it in your own product — no strings.