AI Setup Guide

Set up AI-powered natural language reminders. Type "Call mom tomorrow at 3pm" and GlassNudge figures out the rest.

How it works: You bring your own API key from an AI provider. GlassNudge sends your reminder text to that provider, which returns a structured title and due date. Your API key stays on your Mac—it is never sent to us.

Privacy pick: Cloud providers (Gemini, OpenAI, Anthropic, etc.) send your reminder text to their servers over HTTPS. Ollama and LM Studio run models entirely on your Mac—nothing leaves your machine, and they work offline. If you care most about privacy, start there.

Providers on this page
  1. Google Gemini — best starting point
  2. OpenAI — most widely used
  3. Anthropic — Claude Haiku 4.5
  4. Groq — extremely fast
  5. DeepSeek — great at structured extraction
  6. OpenRouter — access many models through one key
  7. Ollama — local & private
  8. LM Studio — local & private
  9. Custom / other providers

1. Google Gemini

Gemini 3.6 Flash

Model: gemini-3.6-flash · Base URL: https://generativelanguage.googleapis.com/v1beta/openai

Google's fast, capable model and a great default for most people.

  1. Go to Google AI Studio.
  2. Sign in with your Google account.
  3. Click Create API key. Copy the key (it starts with AIza).
  4. In GlassNudge Settings, select Google Gemini from the Provider dropdown—the URL and model fill in automatically.
  5. Paste your key into the API Key field and toggle AI Assistant On.

2. OpenAI

GPT-4o mini

Model: gpt-4o-mini · Base URL: https://api.openai.com/v1

Fast and accurate for reminder parsing.

  1. Go to OpenAI Platform → API Keys.
  2. Sign up or log in.
  3. Click Create new secret key. Copy the key (it starts with sk-).
  4. In GlassNudge Settings, select OpenAI from the Provider dropdown—the URL and model fill in automatically.
  5. Paste your key into the API Key field and toggle AI Assistant On.

3. Anthropic

Claude Haiku 4.5

Model: claude-haiku-4-5 · Base URL: https://api.anthropic.com/v1

Anthropic's fastest model. Supports an OpenAI-compatible endpoint so no special setup is needed.

  1. Go to Anthropic Console → API Keys.
  2. Sign up or log in.
  3. Click Create Key. Copy the key (it starts with sk-ant-).
  4. In GlassNudge Settings, select Anthropic from the Provider dropdown—the URL and model fill in automatically.
  5. Paste your key into the API Key field and toggle AI Assistant On.

4. Groq

Llama 3.3 70B (hosted by Groq)

Model: llama-3.3-70b-versatile · Base URL: https://api.groq.com/openai/v1

Extremely fast inference on Groq's custom hardware. A great pick if speed matters most to you.

  1. Go to Groq Console → API Keys.
  2. Sign up with an email or GitHub account.
  3. Click Create API Key. Copy the key (it starts with gsk_).
  4. In GlassNudge Settings, select Groq from the Provider dropdown—the URL and model fill in automatically.
  5. Paste your key into the API Key field and toggle AI Assistant On.

5. DeepSeek

DeepSeek V4 Flash

Model: deepseek-v4-flash · Base URL: https://api.deepseek.com/v1

Excellent accuracy for structured extraction.

  1. Go to DeepSeek Platform → API Keys.
  2. Sign up.
  3. Click Create API key. Copy the key.
  4. In GlassNudge Settings, select DeepSeek from the Provider dropdown—the URL and model fill in automatically.
  5. Paste your key into the API Key field and toggle AI Assistant On.

6. OpenRouter

OpenRouter (multi-provider gateway)

Model: google/gemini-2.5-flash (or any model) · Base URL: https://openrouter.ai/api/v1

Access hundreds of models through a single API key. Useful if you want to try different models without signing up for each provider. You can switch models just by changing the model name in Settings.

  1. Go to OpenRouter → Keys.
  2. Sign up.
  3. Click Create Key. Copy the key.
  4. In GlassNudge Settings, select OpenRouter from the Provider dropdown.
  5. Paste your key. The default model is Gemini 2.5 Flash, but you can change it to any model on the platform.
  6. Toggle AI Assistant On.

7. Ollama Local

Ollama (local, offline)

Model: llama3.2 (or any model you pull) · Base URL: http://localhost:11434/v1

Run models entirely on your Mac. Total privacy, works offline. No account needed, no internet required after the initial model download. For simple reminder parsing, even the 3B model works well.

  1. Download and install Ollama.
  2. Open Terminal and pull a model: ollama pull llama3.2 (3B, fast) or ollama pull llama3.1:8b (more capable).
  3. Make sure Ollama is running (it runs in the menu bar by default).
  4. In GlassNudge Settings, select Ollama from the Provider dropdown—the URL and model fill in automatically.
  5. For the API Key, type ollama (Ollama does not require a real key but the field cannot be empty).
  6. Toggle AI Assistant On.

8. LM Studio Local

LM Studio (local, offline)

Base URL: http://localhost:1234/v1 · Model: depends on what you load

Run models locally with a friendly desktop app. Browse and download models from Hugging Face directly in the app. No terminal needed.

  1. Download and install LM Studio.
  2. Open LM Studio, search for a model (e.g. "Llama 3.2 3B" or "Qwen 2.5 7B"), and download it.
  3. Load the model and click Start Server (the playground/server tab). The local server runs on port 1234 by default.
  4. In GlassNudge Settings, select LM Studio from the Provider dropdown.
  5. Enter your model name (visible in LM Studio's model list) in the Model field.
  6. For the API Key, type lm-studio (LM Studio does not require a real key but the field cannot be empty).
  7. Toggle AI Assistant On.

9. Custom / other providers

Any OpenAI-compatible endpoint

GlassNudge uses the standard /v1/chat/completions API. Any provider that supports this works—Mistral, Together AI, Fireworks, Cerebras, xAI Grok, Azure OpenAI, and many more.

  1. Get an API key and the base URL from your chosen provider.
  2. In GlassNudge Settings, select Custom from the Provider dropdown.
  3. Enter the Base URL, API Key, and Model name manually.
  4. Click Test connection to verify everything works.
  5. Toggle AI Assistant On.

Frequently asked questions

Where is my API key stored?

Your API key is saved in a local settings file at ~/.glassnudge/settings.json on your Mac. It is never sent to GlassNudge servers (we do not have servers that store user data). It is only sent to the AI provider you configured, over an encrypted HTTPS connection, when you submit a reminder.

What data is sent to the AI provider?

Only the reminder text you type. For example, if you type "Call mom tomorrow at 3pm", that exact text is sent to the provider's API. Your existing reminders, settings, and other data are never included.

Can I use this without an internet connection?

Yes. Set up Ollama or LM Studio with a local model. Everything runs on your Mac with no internet required after the initial model download.

What happens if the AI gets it wrong?

If parsing fails or the AI returns something unexpected, GlassNudge falls back to standard behavior: whatever you typed becomes the reminder title, and your selected time preset is used. Nothing is lost.

Do I need Pro to use AI features?

Yes. AI-powered natural language parsing is a Pro feature. You can enable it from GlassNudge Settings.

Something missing? Email contact@glassnudge.app

Back to glassnudge.app