On July 2 I submitted Jane, an AI SEO agent for Shopify, to the Shopify App Store. I built her alone in four weeks. The idea took three months before I wrote a single line. She is not approved yet. That is exactly why I am writing this now, while the outcome is still unknown, because a build-in-public story told after the trophy arrives is just marketing. This is what it actually took.
- I built and submitted a Shopify SEO app solo. Three months of thinking, four weeks of building, submitted July 2, now awaiting review.
- The whole product hangs on one decision. Jane proposes the work, you approve it, she executes it, and every change reverts in one click. Trust first, features second.
- The hardest bugs were the silent ones. A single character that meant yes on one side and nothing on the other cost me an evening. The logs had the answer the whole time.
- The hardest discipline was not technical. It was cutting features to ship, choosing Shopify's design system over the flashier UI I wanted, and knowing when to stop polishing and hit submit.
- Fresh eyes beat late nights every single time. Every problem I forced at midnight, I solved in minutes the next morning.
- Jane is coming soon to the Shopify App Store. Plans from $39 a month, 14-day free trial on everything.
Table of Contents
- Why I Built Her at All
- The Decision Everything Hangs On
- The Decisions Nobody Warns You About
- Design, and Building a Small World
- The Night a Single Character Beat Me
- The Hardest Part Was Stopping
- Submission Day
- Summary
- Frequently Asked Questions
- Related Reading
Why I Built Her at All
I have optimized more than 25,000 products by hand across my own live stores. Not as a case study. As the person who had to do it, product by product, store by store. So when I went looking for a Shopify SEO app to make that work lighter, I knew exactly what I was looking at, and I did not like it.
Every app fell into one of two camps. The first camp audits your store and hands you a report. A list of problems, color coded, with your evening now planned for you. The app found the work. You still do the work. The second camp goes the other way and bulk-rewrites your catalog in one blast, hundreds of changes at once, with weak or no undo. Handing a stranger the keys to every product page I had sweated over was never going to happen.
What I wanted did not exist. An agent you talk to. One that asks permission before touching anything. One where any change, no matter what, can be undone with a single click. For three months I turned that idea over, checking whether the gap was real or whether I was just not finding the app that already did it. Nothing did it. That is a lonely place for a solo founder, because when nothing like your idea exists, there are two explanations, and only one of them is flattering. I decided to trust my own read of the market, and on June 2 I started building.
The Decision Everything Hangs On
Every product has one decision that everything else hangs from. For Jane it is a four-step loop. Propose, approve, execute, revert.
Jane proposes the work in plain language and tells you what it costs before touching anything. You approve with one click. She executes and shows you before and after. And every change she makes can be reverted, one click, no questions. That is the entire product. Everything else is detail.
Here is the thing. This was never a technical decision. It was a trust decision. Merchants are rightly terrified of apps rewriting their catalog. Your product pages are your livelihood, and an app that changes them without asking is not a tool, it is a liability. So I built the whole product around consent and reversibility, and I made those the load-bearing walls instead of features on a list.
I am explaining this openly because the loop is not the moat. Anyone can read this paragraph. The moat is executing it well, and that took every day of those four weeks.
The Decisions Nobody Warns You About
Nobody warns you that most of building an app is not building. It is deciding what not to build.
The biggest cut hurt. There was a whole category of data access that would have unlocked smarter features, and requesting it triggers a slower, deeper review process before launch. I wanted those features. I dropped them from version one anyway, because a smaller app in merchants' hands beats a smarter app stuck in review. Scope discipline over ambition. That one still stings a little, and it was still right.
The second decision was the free trial. The standard playbook says stack incentives, welcome credits, bonuses, anything to hook the signup. I killed all of it for one clean 14-day free trial on every plan. No asterisks, no games. If Jane cannot prove herself in fourteen days of real use, she does not deserve the subscription.
The third was pricing. Three plans, $39, $99, and $249 a month, sized around how much work Jane does for you rather than which features get held hostage. Every plan gets the whole product. The plans differ only in volume.
And the fourth decision leads straight into design.
Design, and Building a Small World
I can build interfaces. I wanted to build a flashy one, something with my own fingerprints all over it, something that looked nothing like every other app in the admin. Instead I committed to Shopify's own design system from day one, because doing so makes Jane eligible for the Built for Shopify badge, and because an app that lives inside Shopify admin should feel like it belongs there. The disciplined choice over the ego choice. Merchants do not want your creativity in their admin. They want competence that feels native.
Halfway through the build I stopped and did a full redesign anyway, but not for flash. I think of it as building a small world. A grouped sidebar that puts everything one glance away. An overview page I privately call Jane's desk, where she greets you, shows you the state of your store, and points at the next win. The goal was for the app to feel like walking into the office of someone who was already working on your store before you arrived.
Design also handed me my most humbling UX lesson. Jane can optimize a collection page itself, and she can optimize the products inside a collection. Two completely different jobs. My first interface treated them as one concept, and it confused everyone, including me, the person who built it. The fix was not clever. It was two explicit, separately labeled buttons. Clarity beat elegance.
Embedded apps have their own physics too. An app living inside another platform's admin gets reset in ways a standalone app never does. Navigate away, come back, and everything you typed is gone unless you engineer it not to be. And a sophisticated responsive layout component I was relying on simply did not work in that environment. It failed silently, no error, just wrong. I ripped it out and replaced it with something plain and boring that works everywhere. Simplicity over cleverness, again. That theme kept winning all month.
The Night a Single Character Beat Me
Now the war story. One evening, near the end, a feature worked perfectly except that it did nothing. The button clicked. No error appeared. Nothing happened. For an entire evening.
The cause, once I finally found it, was one character. A value crossed a boundary inside the app, a boundary I built myself, as the number 1. The other side would only accept the word true. A strict comparison looked at the number, decided it was not the word, and silently threw it away. Both sides of that conversation were written by me, and they still disagreed about what yes meant.
The bug is not the lesson. The lesson is how long it took me to find it, because I spent that evening guessing. Changing things, redeploying, testing, guessing again. Sometimes testing against a version that had not even finished deploying, which meant my fix was invisible and I concluded it had failed. The moment I stopped theorizing and actually read the logs, the answer was one line. Instant. It had been sitting there the entire time, patiently waiting for me to stop being clever.
There was more of this. I burned real time one exhausted night fighting a cosmetic label on my demo products before recording the submission video, only to discover the label did not block anything at all. I was chasing a gnat in the dark. And in the final stretch, after everything finally worked, the app died on every single message and I was certain I had broken something deep. I had not. The account that powers the intelligence had simply run out of credit. Nothing to do with my code. The log said so, plainly, in one line.
Silent failures are the expensive ones, and it is never what you think. Read the logs. Let the system tell you what is wrong instead of telling the system what you assume. Trusting the evidence over my own theories saved me more hours in four weeks than any tool I used.
The Hardest Part Was Stopping
No bug was the hardest part. Stopping was the hardest part.
When the app was genuinely ready, I circled the finish line for a full day. One more thing to polish. One more wording to tighten. One more edge case to imagine. Every one of them felt urgent and every one of them was procrastination wearing a work costume, because as long as I was fixing things, I did not have to hit submit and let strangers judge four weeks of my life.
Solo is the amplifier here. There is no teammate to say it is done, ship it. You have to be that voice for yourself, and it is much harder to trust that voice when it is the only one in the room. Eventually I recorded the demo video, did the final checks, and made myself stop.
One more thing I learned about that voice. It lies at night. Every problem I tried to force while frustrated and tired went worse, and every one of them fell in minutes the next morning. Rest is not a break from the work. Rest is part of the work.
Submission Day
July 2. Every automated check green. Authentication, security, compliance, all of it. I ran a full self-review on top: 22 checks passing, 0 failing. Then I hit submit, and the strangest phase of the whole project began. The quiet.
Four weeks of motion, of logs and deploys and small victories at strange hours, and then nothing. Somewhere a human reviewer will open Jane, talk to her, click her buttons, and decide. There is nothing left for me to do but write this, and I would rather publish it now, verdict unknown, than dress it up afterward as a story where I always knew it would work. I did not always know. I decided to trust my own judgment anyway, about the gap in the market, about the trust loop, about the boring choice over the clever one, about when it was done. That decision, made over and over, is the actual product of these four weeks.
Jane is coming soon to the Shopify App Store. Three plans from $39 a month, a 14-day free trial on all of them, and a promise built into her bones: she asks first, and everything can be undone.
Meet Jane and get notified when she launches
Summary
- Jane is an AI SEO agent for Shopify. You talk to her, she proposes the work, you approve, she executes, and every change reverts in one click.
- The idea came from optimizing 25,000+ products by hand and finding every existing app was either an audit that dumped homework on me or a bulk-blaster I could not trust.
- Three months of thinking, four weeks of solo building, submitted to the Shopify App Store July 2, now awaiting review.
- The defining decisions were about trust and discipline, not code. Consent and reversibility as load-bearing walls, features cut to ship version one, a clean 14-day trial, and Shopify's native design system over a flashier UI.
- The expensive bugs were silent ones, and the logs answered every single mystery once I actually read them instead of guessing.
- The hardest skill of the whole build was knowing when to stop polishing and submit.
- Jane launches soon. Plans from $39 a month with a 14-day free trial.
Frequently Asked Questions
What is Jane SEO, in plain terms?
An AI SEO agent that lives inside Shopify admin. You tell her what you want in plain language. She proposes the exact work and its cost, you approve with one click, she rewrites your product and collection pages, and any change can be undone with a single click.
How is Jane different from the SEO app I already have?
Most SEO apps either hand you an audit report and leave the work to you, or bulk-rewrite your catalog with changes you cannot easily undo. Jane is a third thing. She does the work herself, one approved step at a time, and everything is reversible. Nothing changes without your yes.
Can Jane break my store?
No. She asks minimal permissions, she never changes anything without your approval, she shows you before and after for every edit, and every change reverts in one click.
How long did it really take to build?
Four weeks of building, June 2 to July 2, as a solo founder. But the honest answer includes the three months before that, turning the idea over and checking whether the gap in the market was real. It was.
When can I get Jane?
She was submitted to the Shopify App Store on July 2 and is awaiting review. The moment she is approved, the listing goes live with three plans from $39 a month and a 14-day free trial on all of them. Details and updates at the Jane SEO page.
Leave a comment (all fields required)