We Tried the AI Agent Route First. That Is Why We Built a Service.

Two price steps from costing more than our biggest plan

The pricing rule is published, and it deserves credit for being published: "After every 20 completed sales the price goes up by 20 %; the counter comes from real transactions, not from a timer." On 24 September their card showed six sales left at the current step.

That step is €119.98, less a €15 discount at checkout: €104.98, or about $120. It is already more than our 100-directory plan and our 30-directory plan together, which is $108 for 130 directories we submit for you.

The next step is €143.98, about $164, which is more than our All Directories tier: $149 once for every directory in our catalog and every one we add, on one project. If the €15 discount carries over, that step lands at about $147, two dollars under it. The step after, twenty-six sales from the day we looked, clears it either way.

And none of those figures includes the agent. Their pricing page: "Your AI provider’s plan limits or metered usage apply." Their blog, on what a run costs in tokens: "I do not know your number, and I will not print a made-up one." That is honest. It is also a bill with no number on it.

39 of their 134 zero-cost routes need your production site

This is the part we care about most, because it is the reason customers come to us. LaunchRepo publishes a table of its 134 zero-cost routes, and in 39 of them the free listing is conditional on a badge or a link on your own website. More than twenty say it has to stay there. One reads, "Badge must stay online; rechecked weekly, 7 days to restore before the listing is removed".

So what does an agent do with a badge? LaunchRepo's answer is honest, and it is the only one a toolkit has. During onboarding the agent asks whether it may "add an official directory badge to that product’s footer, commit, push and redeploy". Their glossary describes the result better than we could: "Adding one means a code change, a commit, a deploy and a public footer that now advertises a third party."

Then there are the routes where the badge has to point at a listing that does not exist until you are halfway through the form. LaunchClash, from their table: "Badge linking to the exact listing URL https://launchclash.com/product/<slug> must be on the product homepage or footer and pass Verify Badge before the Details step unlocks". The form waits while your site ships.

And that is the branch where you said yes. Their blog describes the other one: "A CAPTCHA, a two-factor prompt, an unclear account state or a required badge stops the run", and then "you finish that step in your own browser." So every badge directory ends one of two ways: an agent editing your production site, or you doing it by hand. Neither is paying and leaning back.

That is the pseudo-saving. A toolkit's price stops at the receipt, and every badge adds to it: a commit to production, a deploy, a wait, a re-run and a snippet that has to survive every redesign after it.

We built the other answer. Submitator installs one badge component once and keeps every directory's badge live. No commit per directory, no deploy in the middle of a form and no row of per-directory snippets for next year's redesign to quietly delete. A toolkit cannot ship that, because what keeps it working lives on our side rather than in your repository.

Every CAPTCHA and every code comes back to you

LaunchRepo is precise about the division of labour, so we will quote it rather than paraphrase: "You: Decide which platforms and claims to use; Log in, verify accounts, solve CAPTCHA and 2FA; Grant or withhold submission permission." Codes, their how-it-works page adds, "arrive in your mailbox in the connected browser."

Their own table shows how often that happens. Most of the 77 conditional routes need an account or a sign-in, nine accept Google sign-in only and six put a CAPTCHA, a math challenge or a bot check in the way. Some care about your images down to the kilobyte: ProductFame wants a "logo under 100KB, product image under 500KB", and crozdesk "1-6 screenshots min 1480x987".

None of that surprised us, because it is what we ran into ourselves. A directory took our agent anywhere from five to twenty minutes depending on how many steps its form had. Screenshots had to be cut to each site's size and weight rules in a terminal. And plenty of directories are built badly, with hidden elements and honeypot fields that were hard going even for a capable agent. That is exactly the work we took off the agent and built into our own automation.

Their blog puts the human cost honestly: the agent "needs you within reach during a batch for handoffs", and if you have never run one with a browser, "the first hour is learning, not launching."

Ten at a time, and you watch every run

A service takes one order. A toolkit takes sessions. LaunchRepo's runbook starts you at "a target of 10 new confirmed submissions and at most 2 in parallel". Asked whether Codex can run the whole catalog in one go, their answer is "The runbook says no: small batches". Asked whether it submits while you sleep: "No. Your agent runs in your session, in a browser you can see".

That is the right design for an agent, and we learned why the hard way. On long runs our agent's context filled up and got compacted, and somewhere past the tenth site it started losing the thread. LaunchRepo's answer is a resume file and a saved prompt for the next session. Ours was to take the agent out of the loop altogether.

And watching is the job, not a formality. Their pricing page describes the buyer it fits as someone who can "supervise your agent". Their FAQ explains why that matters: the permission that holds a submission back until you have read it is one of the "instructions for your agent, rather than a software-enforced approval gate." What stands between a draft and a live listing is a sentence the agent has been asked to respect, and the person checking that it does is you.

So a hundred directories through a toolkit, at its starting target, is ten runs you start, supervise and resume, a second job for the length of your launch. Through Submitator it is one payment and a list you tick, and then you lean back while each directory moves from submitted to approved in your dashboard.

Their terms say the guides may go stale

Directories change their steps, their design and their flow all the time. A guide written on the 13th describes the form as it was on the 13th. LaunchRepo is open about this in its terms: "Playbooks describe what we observed on the noted date; we are not obliged to keep them current." And a little further on: "a change made by such a platform after your purchase is not a defect of LaunchRepo." Updates, when they come, are files you copy into your own workspace.

To be fair to them, the changelog is busy: eleven entries between 13 and 23 September 2026 and 165 guides rechecked on 23 September. The question the terms leave open is month six, when a form changes on a Tuesday. With a toolkit, your agent meets the change at the form and, in their FAQ's words, "may need to adapt the guide or pause at a blocker", in your session and on your tokens. With us it is our problem, and we fix it, because we have a steady flow of customers and a changed form is one we are filing ourselves.

Your inbox, or ours

A toolkit signs up with your identity, so it signs up with your mailbox. Their table says as much for F6S: "a 6-digit code emailed on first login; the owner's mailbox is needed". At SaaS AI Tools, "joining subscribes the address to the SaaS Gems newsletter", and more than forty of the 134 routes tell the agent to decline a paid upsell. Every one of those directories now has your address, and they write.

We register you from a dedicated inbox we create for you, and we handle the notifications that land in it. The upsells, the spam and the promotions come to us. Your own inbox stays at zero.

Their agent stops before paying. Our $29 plan already includes the paid tier.

LaunchRepo's workflow "pauses before paid actions and never authorises a purchase for you", which is the right rule for an agent near your card. It also means a toolkit run ends in the free queue, and their own table shows what that can look like: the free queue at Huzzler was "estimated 72 days in Sept 2026".

We pay for the paid tier, or have an arrangement with the directory, so our plans include it: nine directories where even the $29 plan gets the paid tier, worth over $140 at those directories' own prices, with nothing added at checkout. Two of the nine are directories we own, and the savings table says so.

ShowMySites is on both lists, which makes the point neatly. Their route there is the free one. Ours is the $10 Premium tier: a dofollow link and no badge on your site.

What they get right

A lot, and this would be a worse page if we pretended otherwise. LaunchRepo is the most candid product in this category about its own limits. Its terms say it is not a service, its pricing page says who it is not for and its founder writes that he would talk you out of buying if you want somebody else to do the submitting.

Every row in the table carries the date it was checked. The catalog is built on the rule "Unknown is not free." The agent is told to skip a badge directory rather than fake the badge, and never to count a draft as a launch. The company behind it, The CodeCave GmbH, publishes a full imprint, and the licence is generous: one purchase covers every product you own.

Which one should you buy?

Their own blog draws the line close to where we would: "If you ship one product a year, the service column wins on effort and the price difference is small. If you ship several, or relaunch after a rename, the repository column wins on both".

We agree with the first sentence, except that with us the price difference runs the other way: for one product, a hundred directories from us is the cheaper column. On the second, the licence repeats for free and the work does not. Every product is its own run, its own badge deploys, its own handoffs and its own token bill.

If your time is cheap and you already live in Claude Code, their approach works, and it is well documented. If money is the constraint, our smallest plan costs about a quarter of theirs. For everyone else it is a false economy: the receipt is only the first cost, and the rest is paid in your week.

And if the appeal is doing everything at once, our All Directories tier is $149 once for one project: every directory in our catalog (160 today) and every one we add, with nobody asked to supervise a browser.

LaunchRepo alternative: the short answer

LaunchRepo sells good instructions for doing it yourself with an agent, at €104.98 and rising 20% every 20 sales, with your tokens, your logins, your codes and your production deploys on top. Submitator does it for you from $29, badges included.