A lot of AI products feel like they were priced by a VC spreadsheet.
You land on the site, see a polished gradient, and then get hit with $20/month, $30/month, or "contact sales" for something that basically solves one sharp pain.
I didn't want to do that with TokenBar.
TokenBar is a tiny Mac app that sits in your menu bar and shows your token usage and cost in real time while you use LLM apps. I built it because I was tired of that gross feeling of using Claude or ChatGPT heavily and having no idea what the damage would be until later.
The obvious startup advice would have been: make it a subscription.
Recurring revenue. Better LTV. More "serious" business.
Instead, I priced it at $5 lifetime.
That sounds stupid if you look at it like a SaaS founder.
It made perfect sense once I looked at the actual problem.
I wasn't building a platform. I was building a seatbelt.
A seatbelt doesn't need a monthly plan.
The job of TokenBar is simple: sit there, stay out of the way, and stop you from getting surprised. If I solve that cleanly, asking people to rent it forever feels kind of gross.
That realization was useful for me because solo devs can talk themselves into fake sophistication.
I definitely did.
When you're building alone, it's easy to convince yourself the business model needs to be clever because the product is small. So you add tiers, credits, "pro" plans, and a pricing page with enough boxes to look legitimate.
But small products are allowed to be small.
In fact, that's usually the whole point.
The more I worked on TokenBar, the more I realized the real value wasn't in creating some massive AI dashboard. The value was reducing one specific anxiety loop:
"Am I burning way more money than I think right now?"
If the app kills that feeling instantly, $5 is an easy yes.
If it doesn't, no pricing strategy in the world saves it.
That also changed how I shipped the product.
Early on, I was still thinking like I needed to justify a bigger price. I kept wanting to add more. More views. More data. More features that sounded impressive in a screenshot.
Most of that was me avoiding the uncomfortable question: does the core thing already work well enough to charge for?
That's a useful question because it forces honesty.
A lot of builder mistakes are just procrastination wearing a product hat.
You tell yourself you're polishing, but really you're delaying the moment the market gets a vote.
For TokenBar, the biggest lesson was that pricing is product design.
If I charge $5 once, I'm making a promise:
- this app should be simple
- it should keep working
- it should save you more than $5 worth of mistakes almost immediately
- it shouldn't become a needy little business model machine
That promise affects everything.
It affects how much complexity I allow in the UI.
It affects what features I say no to.
It affects whether I build for retention theater or actual usefulness.
And honestly, it makes building more fun.
I'm not trying to trap someone into another monthly expense. I'm trying to make a Mac utility that's so obviously useful that a developer buys it in 30 seconds and forgets the purchase ever happened.
That's the kind of software I like.
The other thing I learned is that shipping paid Mac apps as a solo dev is weirdly clarifying.
Nobody cares how hard it was to build.
Nobody cares that you're tired.
Nobody cares that the implementation was elegant.
They care about one thing:
Does this solve an annoying problem for me right now?
If yes, they'll buy.
If no, they won't.
That's brutal, but it's clean.
So yeah, maybe I could have tried to maximize revenue per user and wrapped TokenBar in a bigger, more expensive story.
But I think there is still room for sharp little software products that do one job well and charge once.
Especially in AI, where everyone is getting fatigued by usage-based pricing, credits, and surprise bills.
Sometimes the better business move is being the product that feels fair.
That's what I'm trying to do with TokenBar.
We'll see if the market agrees.
Top comments (0)