Help with Quick Ask

It's just me making Quick Ask, so the best way to reach me is email: quickask@mikecann.app. Let me know what happened, which iPhone you've got and whether you're using OpenRouter or your own server, and I'll get back to you as soon as I can. A screenshot helps loads, and if it's about speed, tap the timing line under the answer first so the breakdown's in it.

How do I get an OpenRouter key, and what does it cost?

Make an account at openrouter.ai, add a few dollars of credit, then make a key at openrouter.ai/keys. In Quick Ask, tap Paste under OpenRouter key (if you made it on your Mac, Universal Clipboard brings it across). It's kept in your iPhone's Keychain and never shown again, and you can change or remove it in Settings.

Quick Ask itself is free, and you pay OpenRouter directly for what you use (none of it comes to me). A question works out at about a twentieth of a US cent, so a dollar of credit lasts well over a thousand questions, and if you'd like a cap you can give the key a credit limit on OpenRouter.

Which models does it use?

It asks qwen3.8-27b, running on Cerebras, first. If that hasn't started answering within a fraction of a second, it asks gpt-oss-120b on Groq as well, and whichever one starts answering first wins (the other request gets cancelled). They're there because they're some of the quickest you can get on OpenRouter right now. The answers are kept to a sentence or three, and they can't look things up on the web, so they won't know today's news.

You can ask a different model first in Settings › Advanced › Model. Any OpenRouter model ID works there, with gpt-oss-120b still as the backup.

How do I put it on a button?

On an iPhone with an Action button, go to the iPhone's Settings › Action Button, swipe across to Shortcut, pick Quick Ask and then Ask. Or swipe across to Controls and pick the Quick Ask control.

No Action button? Make a shortcut in the Shortcuts app with Quick Ask's Ask action, then go to Settings › Accessibility › Touch › Back Tap, pick Double Tap and choose that shortcut. You can also add the Quick Ask control to Control Center or the Lock Screen, say "Ask Quick Ask" to Siri, or search "Ask" in Spotlight.

However you open it, it's already listening by the time it's on screen, so you can just start talking.

It cut me off, or it waits too long before answering

Quick Ask decides you've finished when you've been quiet for a moment, and you can change how long that moment is in Quick Ask's Settings › Send after a pause of. It starts at 700 ms. If it keeps cutting you off mid-sentence, make it longer; if it feels like it's waiting around after you've finished, make it shorter. When you stop on a word a question can't really end on, like "the" or "and", it waits a bit longer (1.5 seconds) anyway.

You can always tap anywhere while it's listening to send your question straight away. If it's being slow somewhere else, tap the timing line under the answer and it'll show you where the time went.

What do the error messages mean?

If it says "OpenRouter didn't accept your key", the key's wrong, or it's been deleted or disabled on OpenRouter, so make a new one at openrouter.ai/keys and paste it in Settings. If it says "Your OpenRouter account is out of credit", top it up at openrouter.ai and ask again.

"OpenRouter is limiting how fast this key can ask" means it's had a lot of questions in a short time, so give it a minute. Quick Ask sends a question early while it's checking you've finished, and again if you carry on talking, so one question can count more than once.

"You're offline" or "Couldn't reach OpenRouter" mean your iPhone couldn't get through, so check your Wi-Fi or mobile data. "OpenRouter took too long to answer" and "OpenRouter had a problem" are usually on their end, and trying again in a moment normally sorts it. If an answer stops halfway with "The answer was cut off part way through", the connection dropped while it was coming in.

With your own server the messages talk about the server instead of OpenRouter, and anything it sends back in its error goes on the end in brackets.

Can it answer from my own notes or old chats?

It can if you've got a server that knows about them. That's how I use it, with a server of mine (I call it my agent archive) that holds every conversation I've had with my AI agents, so I can ask "didn't I talk to my agent about…?" and it answers from those, with cards for the threads it found underneath.

To use yours, go to Quick Ask's Settings › Answers from, pick Your own server, then type in its address and paste its token. Quick Ask will check you're happy for your questions to go there before it sends anything. Your server needs to answer two requests, and what they look like is all on quickask.mikecann.app/server.