About Sage AI

We believe you should not have to trade away your privacy to benefit from state-of-the-art AI. Sage exists to make the private option the good option — a local-first AI ecosystem capable enough that developers and organizations never have to send proprietary source code, customer data, or personal information to a third party just to get useful work done.

Every hosted AI assistant asks you to make the same trade: upload your codebase, get help. For a weekend project that is fine. For a company's intellectual property, a client's confidential data, or a regulated codebase, it is a decision with consequences — your code now sits on infrastructure you do not control, governed by policies you did not write and that can change without you. Most developers accept this because the alternative has historically been worse tooling. That is the assumption Sage is built to break.

Sage runs on your hardware by default. Install the CLI with pip install sage-ai-cli, pull a model with sage pull, and your prompts, files, and generated code are processed entirely on your own machine. No network round trip, no third-party retention policy, no telemetry carrying your prompt contents anywhere. This is not a privacy mode you switch on — it is the default path, and everything else is the exception.

When you deliberately choose a cloud-hosted model, that request has to leave your machine to be answered. Before it does, Sage strips identifying fields — user ID, email, session identifiers — and scrubs email addresses and phone numbers from the prompt body. The provider on the other end receives an anonymized request that is not linked to you.

Sage supports more than 200 open-weight models, including the Llama, Gemma, Qwen, Mistral, Phi, DeepSeek-Coder, and Devstral families. Open weights matter for a practical reason as much as a philosophical one: a model you can download, inspect, and run yourself cannot be taken away, silently changed underneath you, or made conditional on a subscription. A model on your disk is yours.

Sage is a command line tool, a web app, iOS and Android apps, and an SMS bridge that lets you fire tasks at your own registered computers from iMessage, Google Messages, or email. It integrates with VS Code, Cursor, JetBrains, Neovim, Sublime Text, Zed, GitHub CLI, Git Hooks, Godot 4, and the Unity Editor. The goal is that Sage fits the way you already work rather than asking you to move into a new environment.

Directly, from people who pay for the product. The Free plan is $0 and gives you browser-only AI with 7 in-browser models. Starter is $19/month, Pro is $49/month, and Premium is $99/month, each with an output token allowance and a published per-1,000,000-token overage rate. Payments run through PayPal, Apple Pay, and Google Pay, so we never store your card details.

We say this plainly because business model determines behaviour. A company funded by subscriptions has an incentive to make the tool good. A company funded by data has an incentive to collect more of it. We are the first kind, and we intend to stay that way.

If every server we run went dark, Sage would keep working on your machine — local models would still answer, files would still be edited, commands would still run. That is not a happy accident of the architecture; it is the point of it. A tool that stops being useful when a company has a bad day was never really yours.

We read what people send us, and concrete accounts of real friction shape what we build next far more than abstract feature requests. Tell us what you were trying to do and where Sage got in the way: support@sageworksai.com