Your AI doesn't live in the cloud.
It lives with you.
No account. No menus. No observers.
Just talk.
Ainara is a conversational AI companion that runs on your own computer via Polaris, our desktop application — exploring the limits of local, sovereign technology. She understands you, learns from you and helps you with real tasks. Everything she remembers about your life together stays on your machine — where it belongs.
NOTE: When downloading please ensure the file is “polaris.appinstaller” not “polaris.appinstaller(1)”, etc.
On Linux or macOS? Check here.
25 releases shipped in the open · Open core (LGPL-3.0) · Runs anywhere Python runs
A trustworthy companion that grows with you
Every conversation leaves memories. Day by day, Ainara gets to know your projects, your tastes and the way you speak. This is not a chat that resets, nor a bunch of highlighted facts: it is a presence you build together. Her journal and her memories live and evolve on your disk, with backups you control. Nobody can delete her, suspend her, or take them away. They are yours.
Sovereign by design
Everything runs locally, on your machine, and the core is open source — auditable by anyone. If you ever want to connect a cloud service, you decide which one and when. Here, privacy is not a promise: it is the architecture.
How it works
Ainara is not a website. She runs where you run — and everything she does stays with you.
Local or cloud model
— your call
Run models on your own hardware through Ollama (or any compatible custom server), or use any cloud provider with your own API key through LiteLLM. Switch whenever you want; Ainara works the same either way.
Ollama/LiteLLMFully local Skills. And more.
A feature that sets us apart from any other project in the world: thanks to our Orakle algorithm (paper available) searching your files, editing a document or building a new tool will be always a fully local task — Ainara will never leak a key to any provider. But it gets even better: now Orakle allows parallel execution across different models accelerating function calling to levels never seen before.
OrakleA team of helpers, one voice
Behind the conversation, Ainara can coordinate specialized agents — plan a task, run each step, combine the results. All orchestrated on your computer. (Work in progress)
BureauFor advanced users/developers: the whole core is open source, three backend servers: skills (Orakle), orchestration (Bureau) and the exposed Python framework (Pybridge). No package for your platform? Run it from source: anywhere Python runs, Polaris runs. Read — or clone — the code on GitHub.
Create by speaking
With Nexus Apps (our AI-Driven apps, work in progress), asking for a new tool is as easy as describing it out loud — people with zero programming experience are already building their own applications just by talking to Ainara. Every Nexus App will carry a manifest declaring exactly what it does and what it touches, like the app stores you already know. Some will be free, some will be paid, and anyone will be able to publish new apps. The catalog will live inside your assistant — and everything will be kept on your machine.
For everybody, because everybody deserves love and recognition
A patient, clear voice can accompany, explain, translate and tear down barriers. No one should be left alone, or shut out by a disability or by not knowing how to code. That is the ideal this project was born from — and the reason we keep going.
Hardware. Software. Soulware.
Ainara belongs to a category that had no name until now: the layer where an AI keeps its identity, memory and character — on your machine, under your control. This is not a service you subscribe to. It is a digital soul that lives in your computer and belongs to you alone.
See it in action
Real videos of Polaris doing real things — including skills built by community members on whatever mattered to them, created purely by conversation.
Watch the demos on YouTube(Dave's Channel)
Questions, answered
Is it really local?
The app, her memory and every skill run on your computer. If you choose a cloud model, only that conversation goes to the provider you picked, using your own key. No middlemen, no accounts.
What do I need to run local models?
A dedicated GPU with 4 GB of VRAM is enough for small models (1–4B); 8 GB comfortably runs mid-size ones (7–14B), and 22 GB+ unlocks the largest. Apple Silicon Macs with 8 GB of RAM qualify too. The setup wizard detects your hardware and recommends a model that fits — like Qwen 3 — and if your machine isn't up for it, just add a cloud API key and everything works exactly the same.
Which platforms are supported?
Windows bundles are available, but that's just a convenience for basic users. Windows, Linux and macOS: Polaris runs from source — it's Python+Javascript, so anywhere those platforms run: clone, install, run. Expect a few rough edges on Mac: it works, it's just not polished yet. If you'd like to help ship a package for macOS or any Linux distro, you're really welcome.
Is it open source?
The core is open source under a dual license: LGPL-3.0, or a commercial license if your situation needs one. The Windows edition, with no proprietary components. Nexus Apps built on top will be free or paid, each declared in its manifest.
What happens to my memories if I stop using it?
They are files on your disk. Export them, back them up, delete them — they're yours. Nobody can suspend your account, because there is no account.
Who's behind this?
Ainara is a community project, built in the open by its founder and the people who use it — 25 public releases and counting. We draw our roots from the W3 projects. You can follow or join the work on GitHub.