Jotley used to have subscription tiers metered by minutes and summaries. Today it's free. That change came straight out of the architecture. I moved the whole app onto the device, the per-minute bill went away, and I could no longer say what a monthly fee was paying for.
I just added optional tips to Jotley and PulseCanvas Studio. Both apps stay free, and a tip unlocks no features. The reasoning applies to anyone deciding how to charge for an app right now: a mandatory monthly fee needs a monthly cost behind it. If you can't point at the bill, I don't think you can defend the fee, and agents are making that harder.
What a real bill looks like
Old Jotley was a normal SaaS. Every recording went up to my API, Deepgram transcribed it, the OpenAI API summarized it and iCloud kept the result.
The shape of that cost is easy to show. Deepgram's current list price for Nova-3 pre-recorded transcription is $0.0043 a minute on pay as you go. Someone who records 20 hours of meetings a month uses 1,200 minutes, which comes to $5.16 for transcription alone, before summaries or storage. Someone who records 40 hours costs twice that. That's today's price and a made-up user, so don't read it as a Jotley number.
The meter existed because the invoice did. A flat unlimited plan would have let my heaviest users cost the most and pay the same as everyone else. A subscription was fair there. Users were covering their share of a cost that showed up every month and grew with every minute they recorded.
What the bill looks like now
Jotley now transcribes with Apple's speech models and summarizes with Apple Intelligence on the user's own device. Apple pitches its Foundation Models framework to developers as "AI inference that is free of cost." The compute is hardware the user already bought. If iCloud sync is on, the text of their notes moves through their own private iCloud. One more hour of recording costs me nothing.
PulseCanvas Studio never had a cloud pipeline to remove. The user's GPU does the rendering and their music stays on their device.
What's left is a short list. The Apple Developer Program is 99 USD a year. PulseCanvas Radio needs storage I run: its songs and station lists sit in a Cloudflare R2 bucket as public static files, with no API in front of them. R2 bills for storage and operations and charges nothing for egress, so Radio's cost looks nothing like paying for every minute of someone's meeting.
Here's the test I'd run on any app. Imagine you stop paying every vendor tomorrow, then ask what breaks on the user's device.
Old Jotley stops transcribing and summarizing. Current Jotley keeps working, and PulseCanvas Studio only loses Radio.
If nothing important breaks, a required monthly fee is rent on something the user already has.
Agents make the fee harder to defend
The usual defense of a subscription on a no-server app is development: the fee pays for updates. Apple is fine with that argument. Its App Review Guidelines require a subscription to "provide ongoing value to the customer" and list "apps that offer consistent, substantive updates" as an example. So the bar I'm describing is my own, and no App Store rule backs it up.
That defense depends on development staying expensive, and that's the part that's moving. PulseCanvas went from 31 designs to 336 in about a week with agents writing the shaders. Agents wrote Jotley's Swift too. What cost me something was the decisions, like tearing out a cloud pipeline that worked and cutting visuals that had passed every test.
Run that forward. If an app has no server and its code keeps getting cheaper to write, someone with agents can build a close copy and give it away, because running it costs them almost nothing either. A required monthly fee with only code behind it is about the easiest price in software to undercut. I expect a lot of small utility apps to run into this.
Real SaaS is in a different spot. Agents lower what it costs to write the software, and they do nothing to the invoice from your transcription vendor or your database host. If you run models in the cloud or hold people's data for them, that cost shows up every month whether you ship anything or not. A monthly price is the right way to cover it.
Where this argument is weak
Maintenance isn't free. Jotley is built on Apple's on-device models, so when Apple changes them the app has to keep up, and every prompt change now ships as an app update. Agents can do that work, but someone still has to decide what's worth changing, and that time recurs too. The difference is that a no-server app still runs in a month when nobody ships an update. Old Jotley couldn't run in a month when nobody paid for transcription.
That recurring time is what the tip screen asks people to back: "If you'd like to support future updates, you can make a one-time contribution." Asking is fair. Making it the price of opening the app is where I get off.
Tips may also not add up to much. They're brand new, but with only a small number of downloads so far, I've already gotten a couple of tips. That's a good early sign, and it's still too soon to tell you whether this works as a business. If a no-server app has to carry a whole team, you have a harder decision than I do. I'd look at a one-time price first, since it matches a cost you paid once.
Free also doesn't make an app easy to find. I tried running ads for Jotley, and every click cost about $2, because big SaaS companies market AI transcription so aggressively. Needless to say, that's not sustainable for an app I'm giving away. A free app has no marketing budget built into its price, so Jotley has to grow by word of mouth.
Match the price to the cost
The rule I'd use fits in a code block.
How your cost behaves What you can fairly require
Built once, runs on their device Nothing, or a one-time price
Flat hosting every month A flat monthly fee
Per minute or per token Metered tiers or credits
Flat hosting plus usage A base fee plus usageOn iOS, guideline 3.1.1 explicitly allows customers to tip a developer through in-app purchase. Apple's commission applies, and it's 15% under the Small Business Program if you qualify. Mine are consumable purchases in three one-time sizes from $0.99 to $5.99. The comment above the tip code is the whole policy: "Consumable purchases are a thank-you, never a feature entitlement."
Before you put a required monthly fee on an app, write down the bill it pays for. If you can't name one, charge once, or ship it free and ask for a tip.
If you think I've got this wrong for your app, tell me what your bill looks like.