> ## Documentation Index
> Fetch the complete documentation index at: https://docs.wolffi.sh/llms.txt
> Use this file to discover all available pages before exploring further.

# Installation

> Install Wolffish on macOS, Windows, or Linux

# No Dependencies, Just Install

Wolffish is a self-contained Electron app. No Node.js, no Python, no Docker. Just install and run.

## Prerequisites

<AccordionGroup>
  <Accordion title="LLM Provider (at least one required)" defaultOpen>
    Wolffish needs at least one LLM provider to think and respond. You can use **cloud providers**, **local models via Ollama**, or both:

    * **DeepSeek (V4 Pro)** ⭐ Recommended — Get an API key at [platform.deepseek.com](https://platform.deepseek.com)
    * **Anthropic (Claude)**: Get an API key at [console.anthropic.com](https://console.anthropic.com)
    * **OpenAI (GPT)**: Get an API key at [platform.openai.com](https://platform.openai.com)
    * **Ollama (Local)**: Install from [ollama.com](https://ollama.com) for local model inference

    You'll configure these in Wolffish's settings after installation. You explicitly select your Brain model — a cloud provider or local Ollama — and that's the one that runs. There's no automatic fallback between them. You can use only cloud providers, or run fully local by selecting Ollama as your Brain.
  </Accordion>

  <Accordion title="Ollama (Optional — powers local models)">
    Ollama lets you run open-source models entirely on your machine with zero internet dependency. Wolffish integrates directly with it and will help you choose and download a model. No terminal commands needed.

    Ollama is **not required** — you can use Wolffish with only cloud providers (DeepSeek, Claude, OpenAI). But if you want offline capability or prefer to keep all inference local, install Ollama.

    <Note>
      Reliable agentic tasks (multi-step tool use) require large models — 70B+ parameters (Llama 3 70B, Qwen 2.5 72B, DeepSeek-V2). Running these needs 48GB+ RAM or a high-end GPU. Smaller models (7B–14B) work for simple conversations but struggle with complex tool-calling workflows. See the [Ollama integration guide](/configuration/ollama) for details.
    </Note>
  </Accordion>

  <Accordion title="macOS Permissions">
    For the computer-use capability, macOS will prompt you to grant **Accessibility** and **Screen Recording** permissions. These are only needed if you want Wolffish to interact with your screen.
  </Accordion>
</AccordionGroup>

## Download

<Tabs>
  <Tab title="macOS">
    Download the latest `.dmg` from the [download page](https://www.wolffi.sh/).

    1. Open the `.dmg` file
    2. Drag Wolffish to your Applications folder
    3. Launch Wolffish from Applications

    <Note>
      On first launch, macOS may warn about an unidentified developer. Right-click the app and select "Open" to bypass this.
    </Note>
  </Tab>

  <Tab title="Windows">
    Download the latest `.exe` installer from the [download page](https://www.wolffi.sh/).

    1. Run the NSIS installer
    2. Follow the setup wizard
    3. Launch Wolffish from the Start menu

    All permissions are auto-granted on Windows.
  </Tab>

  <Tab title="Linux">
    Download the latest `.AppImage` from the [download page](https://www.wolffi.sh/).

    ```bash theme={null}
    chmod +x Wolffish-*.AppImage
    ./Wolffish-*.AppImage
    ```

    <Warning>
      X11 is required. Wayland is not currently supported for the computer-use capability.
    </Warning>
  </Tab>
</Tabs>

## What's Bundled

You don't need to install anything else. Electron bundles Node.js and Chromium internally. The app ships with:

* Runtime environment (Node.js + Chromium)
* All 20+ built-in capabilities
* SQLite (FTS5 search index)
* All channel adapters (Desktop, Telegram, WhatsApp)

## Requirements by OS

|                     | macOS                            | Windows                 | Linux                           |
| ------------------- | -------------------------------- | ----------------------- | ------------------------------- |
| **Install method**  | `.dmg` → drag to Applications    | NSIS installer (`.exe`) | AppImage (mark executable, run) |
| **Node.js needed?** | No                               | No                      | No                              |
| **Runtime deps**    | None                             | None                    | None                            |
| **For local AI**    | Ollama (optional)                | Ollama (optional)       | Ollama (optional)               |
| **For cloud AI**    | API key (Anthropic or OpenAI)    | API key                 | API key                         |
| **Permissions**     | Accessibility + Screen Recording | Auto-granted            | X11 required                    |

System tools like `ffmpeg` and `git` are **not prerequisites**. When a capability needs one, Wolffish detects it's missing and offers to install it for you (via your system's package manager), with your approval. The agent handles dependency management at runtime.

## Build from Source

<Warning>
  Building from source is only for contributors. Regular users should use the
  pre-built installers above.
</Warning>

```bash theme={null}
git clone https://github.com/thewolffish/wolffish-app.git
cd wolffish-app
npm install
npm run dev
```

Building from source requires Node.js 20+ and npm. Native modules (better-sqlite3) are compiled during `npm install` via `electron-rebuild`.

## Verify Installation

On first launch, Wolffish creates `~/.wolffish/workspace/` with default configuration files. The onboarding flow will guide you through choosing a model and getting started.

To verify everything is working:

1. Follow the onboarding prompts to select and download a model
2. Send a test message like "Hello, what can you do?"
3. You should see a streaming response

If Ollama is not detected, Wolffish will offer to guide you through installing it — or you can skip and use cloud providers instead.

## Uninstall

All user data lives in `~/.wolffish/`. To fully uninstall:

1. Delete the application
2. Remove the data directory: `rm -rf ~/.wolffish/`