Bring your own key
wuup's free tier gives you 50 questions on Gemini. To keep going, or to use a different model, add your own API key from OpenAI, Anthropic, OpenRouter, Google Gemini or Groq in Settings. With your key, questions are unlimited on the wuup side and wuup charges nothing: your provider bills you directly for the tokens you use. wuup checks the key when you save it, shows it masked afterwards, and you can remove it at any time.
Supported providers
| Provider | Where to create a key | Notes |
|---|---|---|
| OpenAI | platform.openai.com/api-keys | Needs billing set up on your OpenAI account |
| Anthropic | console.anthropic.com/settings/keys | Claude models, billed by Anthropic |
| OpenRouter | openrouter.ai/keys | One key for many model vendors, prepaid credits |
| Gemini | aistudio.google.com/app/apikey | Google AI Studio key |
| Groq | console.groq.com/keys | Very fast open models, keys start with gsk_ |
Each provider shows the full secret only once, when you create it. Copy it straight into wuup or a password manager.
Add your key
- In wuup, open Settings, then API key.
- Choose the provider.
- Paste the key and choose Save.
wuup validates the key by asking the provider for its model list. Listing models does not spend tokens. If the provider accepts the key, wuup saves it, stores the list of models your key can use, and the model picker in the chat switches to those models. If the provider rejects it, nothing is saved and you see an error such as "Invalid Anthropic API key". See Troubleshooting if that happens.
Choose a model
After you save a key, the model picker lists the models your key can access. Pick one per chat, or leave it on the default:
- OpenAI defaults to
gpt-4o-mini. - Anthropic defaults to a current Claude model.
- OpenRouter defaults to
google/gemini-2.5-flash. - Gemini uses wuup's rotation of current Gemini models.
- Groq picks the strongest tool-calling model your key can use (see Groq below).
Every answer records which model produced it and whether your own key was used.
Groq
Groq serves open models, such as Llama and GPT-OSS, on its own inference hardware, so answers stream back very fast. To use it:
- Sign in to the GroqCloud console and create an API key. It starts with
gsk_. - In wuup, open Settings, then API key, choose Groq and paste the key.
When you save the key, wuup lists Groq's models with it and keeps only the chat models. Speech, text to speech and safety models are left out, because they can't answer questions about your data. Auto then uses the strongest tool-calling model your key can see, and a small, fast model for chat titles.
Groq applies rate limits per model, and the limits on its free plan are low. If a model is rate limited, Auto moves on to the next model and the model picker shows when the limited one is free again. If you pick a specific model and it hits its limit, wait for the time shown or switch back to Auto. Groq bills by tokens at the rates on groq.com/pricing; wuup adds nothing.
How your key is used
- wuup sends your key only to the provider you chose, to stream the answer and to run tool calls such as database queries.
- After saving, the key is never shown in full again. Settings shows a masked hint with the first four and last four characters, for example
sk-p••••a1b2. - Questions asked with your key do not count against the free tier.
- wuup adds no markup and no per-question fee.
Remove or replace your key
Open Settings, then API key, and choose Remove. wuup deletes the stored key and its model list, and new questions go back to the free tier if you have questions left. To switch providers, remove the old key and add the new one. Rotating the key at your provider also cuts off wuup immediately.
FAQ
Which provider should I pick?
For data questions, any current mid-size model works well. OpenRouter is useful if you want to try several vendors with one key, and Groq is the pick when you want answers back as fast as possible. Our guide BYOK explained compares the trade-offs.
Do I still need a wuup account?
Yes. Your key pays for the model, while wuup runs the connections, the read-only guard, charts and chat history.