Wiki / Connect an AI
Connect an AI
Card Coach AI, free OpenRouter models, or your own key.
Card Coach needs an AI service to answer. You have three choices, and you can switch at any time from Card Coach → Connection. Each provider keeps its own key, so switching back and forth doesn't lose anything.
| Option | Cost | Best for |
|---|---|---|
| Card Coach AI | Free starter quota, then optional top-ups | The easiest setup. No API key. |
| OpenRouter | Free with :free models | Studying for free, or trying many models. |
| Your own key | You pay your provider directly | Kimi, OpenAI or any OpenAI-compatible server. |
Option 1: Card Coach AI
- Open Card Coach → Card Coach AI and click Free. This opens the Connection page with Card Coach AI selected.
- Click Sign in (or Create account). This website opens in your browser.
- Sign in, or create an account and confirm your email. Anki connects automatically. You never type a password or key into Anki.
Once connected, the Connection page shows your email, the quota you have left, and buttons to refresh, top up or sign out.
Card Coach models
- Card Coach: the everyday model. Fast, and the default.
- Card Coach Thinking: reasons step by step before answering. Slower, and uses more quota. You can also switch it on for a single question with the Thinking button in the chat box.
Option 2: OpenRouter (free models)
- Create a free account at openrouter.ai and create an API key. No credit card is needed.
- In Connection, choose OpenRouter and paste the key.
- In Model, click Browse… and pick a model whose name ends in
:free. Free models are shown in bold.
Free OpenRouter models have daily rate limits and come and go. If one stops answering, browse again and pick another.
Option 3: Your own provider
- Kimi (Moonshot): choose Kimi, paste your Moonshot key, then pick a model such as
kimi-k2.6. - Custom: any OpenAI-compatible endpoint, such as OpenAI, a company gateway or a local server. Enter the API base URL, your key, and type the model name.
The model browser
The Browse… button next to the Model field opens a searchable list fetched live from your provider. Each row shows the context length, price per million tokens, and whether the model accepts images. Type to filter; several words must all match. The list is cached for 24 hours, and ↻ Refresh fetches a new one.
The same list is in the model dropdown of the chat panel. It has a search box at the top and a ↻ Refresh latest models row, so you can change models without leaving the card.