Canonical: https://krasis.ai/docs/getting-started/

[Documentation](https://krasis.ai/docs/)

Start here

[Overview](https://krasis.ai/docs/)[Getting started](https://krasis.ai/docs/getting-started/)

Everyday guides

[Choosing a model](https://krasis.ai/docs/models/)[Connecting your apps](https://krasis.ai/docs/clients/)[Memory and speed](https://krasis.ai/docs/memory/)[Troubleshooting](https://krasis.ai/docs/troubleshooting/)

Technical reference

[Configuration](https://krasis.ai/docs/configuration/)[API and authentication](https://krasis.ai/docs/api/)

For Krasis 1.0.22  
[Release history](https://krasis.ai/releases/)

Krasis documentation

# Getting started

Install Krasis, choose a model and get your first reply.

On this page

[1\. Check your machine](https://krasis.ai/docs/getting-started/#1-check-your-machine)[2\. Install Krasis](https://krasis.ai/docs/getting-started/#2-install-krasis)[3\. Choose a model](https://krasis.ai/docs/getting-started/#3-choose-a-model)[4\. Launch with the suggested settings](https://krasis.ai/docs/getting-started/#4-launch-with-the-suggested-settings)[5\. Connect a chat or coding app](https://krasis.ai/docs/getting-started/#5-connect-a-chat-or-coding-app)[6\. Save, manage and update](https://krasis.ai/docs/getting-started/#6-save-manage-and-update)

## 1\. Check your machine

You need an NVIDIA Ampere or newer GPU with a working NVIDIA driver, plus enough system RAM and disk space for your chosen model and its caches. Krasis runs on Windows, Linux and WSL2\. macOS is not supported.

Use the [install page’s hardware checker](https://krasis.ai/install/) before downloading a large model. Choose a smaller model for your first run if memory is tight.

## 2\. Install Krasis

**Windows:** download the current installer from the [install page](https://krasis.ai/install/#windows), run it and open **Krasis** from the Start menu. The installer includes its own Python runtime. The first setup downloads CUDA dependencies and can take several minutes.

**Linux or WSL2:** use Python 3.10–3.13 and run:

```
curl -fsSL https://krasis.ai/downloads/install.sh | bash
krasis-setup
krasis
```

If `krasis-setup` is not found, reopen your terminal and try again. On WSL2, run these commands inside Linux.

## 3\. Choose a model

The launcher offers a supported-model download flow. Select a model that fits your RAM and GPU budget, or select a supported model you already have locally. Krasis uses Hugging Face safetensors checkpoints; a GGUF file is not a substitute for this workflow.

Downloads can be large. See [Choosing a model](https://krasis.ai/docs/models/) before committing disk space to one of the largest options.

## 4\. Launch with the suggested settings

Select your GPU and start with the launcher’s suggested settings. Keep experimental options off for your first run. Select **Launch** and wait for the server to report that it is ready.

The first launch builds optimised model caches and calibrates memory. This can take much longer than later launches. Saved settings live under your user’s `.krasis/config` directory.

## 5\. Connect a chat or coding app

In an OpenAI-compatible app, set the API base URL to:

```
http://127.0.0.1:8012/v1
```

Use the host and port shown by Krasis if you changed them. Select the model reported by the server and send a short prompt such as “Explain what a mixture-of-experts model is in three sentences.”

If API authentication is enabled, enter one of your key values in the app’s API-key field. See [Connecting your apps](https://krasis.ai/docs/clients/) for model discovery and a direct request example.

## 6\. Save, manage and update

Reuse the launcher’s saved profile for future runs. For a browser dashboard, open **Krasis Manager** on Windows or run `krasis manager` on Linux. Manager opens locally at `http://127.0.0.1:8090` by default.

On Linux, update with `krasis update`. On Windows, run the current installer from the install page.

Next: [connect your apps](https://krasis.ai/docs/clients/), [control access with API keys](https://krasis.ai/docs/api/#api-keys), or learn [what the memory settings mean](https://krasis.ai/docs/memory/).