Quick Ask

For when you're out and just want a quick answer

So imagine you're out with a mate and you want a quick answer to something, or you suddenly think "hang on, didn't I talk to my agent about that?" Well that's what I made Quick Ask for. I just press the Action button, ask out loud, and a second or two after I stop talking it starts reading me the answer. Hopefully you find it handy too!

Coming soon to theApp Store

It's free, you just bring your own OpenRouter key.

The answer to How long do I boil an egg for a runny yolk: about six minutes from when it goes into boiling water, then straight into cold water. Under it, 840 ms to first word
Didn't I talk to my agent about what to plant in the veggie patch, answered from your own server: cherry tomatoes and basil along the back fence and lettuce in the shady corner. Under it, 1.28 s to first word and cards for the three old chats the answer came from

How I use it

A quick answer while you're out

Say you're at the pub and someone wonders how far away the Moon actually is, or you're at the shops trying to remember what 350 Fahrenheit is in Celsius. I didn't want to stop and type it in and scroll past a page of results, I just wanted the answer. So you press the button, ask, and it reads it back to you in a sentence or two.

It remembers your last three questions too, so you can tap the big mic and ask a follow-up like "and how long should I rest it for?" and it knows what you're on about.

The answer to How far away is the Moon: on average about 384,000 kilometres. Under it, 790 ms to first word
A follow-up, And how long should I rest it for, under the faded question before it about cooking a two kilo leg of lamb, answered with 15 to 20 minutes loosely covered with foil

"Hang on, didn't I talk to my agent about that?"

This was a big part of why I made it. I've got a server of my own (I call it my agent archive) that holds every conversation I've had with my AI agents, and Quick Ask can ask it. So when I'm out and half remember working something out with my agent a few weeks back, I just ask, and it answers from those old chats and shows me the threads it came from underneath.

My archive's just for me at the moment, but you can point Quick Ask at a server of your own, with your own notes or agent history in it. What your server needs to do is all on this page.

Didn't I talk to my agent about what to plant in the veggie patch, answered from your own server: cherry tomatoes and basil along the back fence and lettuce in the shady corner. Under it, 1.28 s to first word and cards for the three old chats the answer came from

As quick as I could make it

Speed was the whole point, so it's already listening when it opens and you just start talking. The moment you pause it sends your question off, while it's still checking you've actually finished, and then it reads the answer out a sentence at a time as it streams in rather than waiting for the whole thing.

Under every answer there's a little line like "840 ms to first word", which is how long it was from when you stopped talking to the first word you heard. Tap it and you get the whole breakdown of where the time went (I'm a bit obsessed with that number).

350 Fahrenheit in Celsius, with the timing line tapped open to show the breakdown, from press to mic live down to 860 ms from end of speech to the first spoken word
The timing breakdown up close: 210 ms from press to mic live, 302 ms from end of speech to request sent, 308 ms from request to first text from qwen3.8-27b on Cerebras, all of it hidden behind the 700 ms pause, and 860 ms from end of speech to the first spoken word

Where the answers come from

Out of the box there's no server in the middle. You paste in your own OpenRouter key and your iPhone asks the model directly. It asks qwen3.8-27b on Cerebras first, and if that's being a bit slow it asks gpt-oss-120b on Groq as well and goes with whichever one answers first. They're some of the quickest models you can get on OpenRouter, and it works out at about a twentieth of a US cent a question on your own OpenRouter bill, so a few dollars of credit goes a long way.

Or you can point it at your own server, like I do with my archive, and then answers can come with cards for the chats or notes they came from. You can switch between the two in Settings whenever you like.

The first run asking where answers should come from: OpenRouter, using your own OpenRouter key, or your own server, with the OpenRouter key saved

Put it on a button

It's quickest on the Action button (Settings › Action Button › Shortcut › Quick Ask), then it's one press and you're already asking. However you open it, it's listening by the time it's on screen.

If your iPhone doesn't have an Action button you can use Back Tap with a shortcut, add the Quick Ask control to Control Center or the Lock Screen, or just say "Ask Quick Ask" to Siri.

Put it on a button: the Action button, Back Tap, Control Center or Siri, with how to set each one up

Tap anywhere

Tap anywhere while it's listening and it sends your question straight away, or tap while it's talking to make it stop. New up the top starts a fresh conversation.

It keeps going when you lock it

If you lock your phone after asking, it keeps reading the answer out, so you can put it straight back in your pocket.

What gets sent

iOS turns your voice into text, on your iPhone where it can, and only the text of your question goes off to be answered. Quick Ask asks before it sends anything anywhere, and none of it comes to me. The privacy policy has the details.

Give it a go

Quick Ask is free on iPhone, for iOS 18 or later. You'll need an OpenRouter key (or your own server) for it to answer anything. If you find a bug or have an idea for it, drop me an email at quickask@mikecann.app.

Coming soon to theApp Store