Free and open source, MIT licensed

I wanna

wanna.sh runs AI on your own computer. Chat, pictures, video, voices, transcription, music and a coding agent in one app, private and working offline once a model is downloaded.

Get wanna.shRead the docs

Get wanna.sh

On Windows, download the app and open it. It sets everything up the first time, no terminal needed. On Linux and macOS, one command does the same.

What your computer needsStep-by-step installRun in Docker

Download for Windows

Windows 10 or 11, 64-bit. Free and open source.

Linux and macOS

curl -fsSL https://raw.githubusercontent.com/wanna-sh/wanna/main/install.sh | bash

Then run wanna and open http://localhost:7860.

Six AI tools, one app

Instead of a separate app for each kind of AI, wanna.sh brings popular open models together, with the same simple controls everywhere.

Made for people, not just engineers

Every model is described in plain language and rated for your computer before you download it. No terminals, no config files, no guessing.

The Chat page: a conversation with Llama 3.2 3B, with model, attach and instructions controls under the message box.
Get models: each model shows how well it runs on this computer, like Runs great here or Runs well here.
Get models. Thousands of models from Hugging Face, each rated: runs great, runs well, runs slowly, or too big.
Code: an agent searched the web and wrote a file, each step shown as a card.
Code. A coding agent that writes files, runs commands in a Docker sandbox and searches the web, with your approval.

Why run AI on your own computer?

  • Private by design

    Your prompts, chats and creations stay on your computer. The internet is only used to find and download models.

  • Works offline

    Once a model is downloaded it keeps working without a connection, on a plane or behind a firewall.

  • Free and open source

    MIT licensed and developed in the open. No accounts, no subscriptions and no per-use limits.

  • Uses all your hardware

    Models go on the right graphics card automatically, and big ones are shared across several cards.

How it works

  1. Install it

    One command sets up everything on Linux, macOS or Windows (with WSL2). Then open it in your browser.

  2. Pick a model

    Browse thousands of models from Hugging Face, each rated for your computer, so you know it will run before you download.

  3. Say what you want

    Type a request, choose a few simple options and press Create. The model starts by itself the first time you use it.

Built for your other apps too

wanna.sh speaks the OpenAI API. Point any app, script or plugin that supports a custom OpenAI address at wanna.sh, and it uses your own models for chat, pictures, speech, transcription and video.

Connect an appAPI reference

from openai import OpenAI

client = OpenAI(
    base_url="http://localhost:7860/v1",
    api_key="not-needed",
)

reply = client.chat.completions.create(
    model="auto",
    messages=[{"role": "user", "content": "Hello!"}],
)
print(reply.choices[0].message.content)

Your computer is more capable than you think.

Install wanna.sh, find a model that suits your computer, and start making things.

Get startedView on GitHub