Smart Window lets you connect your own AI model instead of using the ones provided by Firefox. This is helpful if you want more control, use a specific provider, or run a model locally on your device.

You can connect either:

• A remote model (such as OpenRouter)
• A local model running on your device (such as Lemonade Server or Ollama)

Use a remote model (OpenRouter)

1. Create an OpenRouter account if you do not have one already, at https://openrouter.ai/.
2. Generate an API Key in OpenRouter, and copy it to a secure place.
   • OpenRouter API keys begin with sk-or-v1-.
3. Open the OpenRouter models page and choose a model you would like to use.
   • Take note of its model ID. Ex: z-ai/glm-4.5-air:free.
4. In Firefox: In the Menu bar at the top of the screen, click Firefox and select Settings (or Preferences, in some cases).Click the menu button and select Settings.
5. Go to AI Controls > Smart Window Settings > Assistant model.
6. Select Custom: Use your own LLM.
7. Fill in the fields:
   • Model name: Paste OpenRouter model ID, from step 3.
   • Model endpoint with the OpenRouter API endpoint, which is typically https://openrouter.ai/api/v1.
   • API key: Paste your OpenRouter API key from step 2.
8. Click Save.
9. Open a Smart Window, and start using the Assistant.

Use a local model

Example: Lemonade Server

1. Download and install Lemonade Server at https://lemonade-server.ai/. You must use version 10.2.0 or newer.
2. Run Lemonade Server and download a model of your choice using the app instructions.
3. In a command line terminal, set a larger context size by using the command lemonade config set ctx_size=8192.
4. Reload the model from the UI or by using the command lemonade unload (the next time you make a request to the model, it will load with your settings).
5. In Firefox: In the Menu bar at the top of the screen, click Firefox and select Settings (or Preferences, in some cases).Click the menu button and select Settings.
6. Go to AI Controls > Smart Window Settings > Assistant model.
7. Select Custom: Use your own LLM.
8. Fill in the fields:
   • Model name: Enter your model name from step 2 (for example, SmolLM3-3B-GGUF).
   • Model endpoint: Enter the Lemonade Server endpoint, which is typically http://localhost:13305/api/v1.
   • Note that no API key is required for Lemonade Server.
9. Click Save.
10. Open a Smart Window, and start using the Assistant.

Example: Ollama

1. Download and install Ollama at https://ollama.com/download.
2. Run Ollama, and follow the instructions on the site to download a local model of your choice.
3. Open the Firefox settings screen, and go to AI Controls > Smart Window Settings > Assistant model, and select Custom: Use your own LLM.
4. Fill in the fields:
   • Model name: Enter your model name from step 2 (ex: qwen3.5:4b).
   • Model endpoint: Enter the Ollama endpoint, which is typically http://localhost:11434/v1.
   • Note that no API key is required for Ollama.
5. Click Save.
6. Open a Smart Window, and start using the Assistant.

Related articles

• Model choices in Smart Window
• Memories in Smart Window
• Get started with Smart Window