Before you start
A submission is a form on somebody else's website, so all of this happens in a browser. Which browser your agent drives, and what you let it do with your accounts and your repository, is most of the difference between a run that finishes and one you end up finishing by hand.
What your agent needs to be able to do
Three things, and only the first is about us. It has to reach this server, which is the setup page. It has to drive a browser, which is the section below. And it helps enormously if it can see the product's own folder, which is the section after that.
Which browser it drives
All four of these work. They differ in how many times the run stops and asks you for something.
The browser you use every day, with an agent extension in it. You are already signed in to the accounts these platforms ask for, the profile has real history behind it, and nothing has to be set up per site.
A second MCP server that drives Chrome, told to launch with your existing profile rather than a blank one. Same sessions as above, and it works in a terminal client that has no browser of its own.
A clean profile with no history and no logins. Every platform that wants an account stops the run, and anti-bot checks fire far more often, because from the outside a fresh automated profile is indistinguishable from a bot. Fine for reading a form, slow for filling one.
The agent moves your actual cursor and types into whatever window is in front. It works with any browser and needs nothing installed in it, but it is the slowest of the four and it can see your whole screen while it runs.
Why your own browser meets fewer captchas
Cloudflare, reCAPTCHA and the rest are not deciding whether you are allowed in. They are scoring how much you look like a person, from the profile's history, the accounts already signed in, the network you are on and the shape of the browser itself. Your everyday Chrome scores like a person because it is one. A fresh automated profile, opened for this and thrown away after, scores like a bot for exactly the same reasons.
That is not a way around a challenge and nothing here tries to be. When one does appear, the agent stops and hands it to you, on any of the four options above. Working in your own browser just means it appears a great deal less often.
Signing in to the platforms
Plenty of these platforms want an account before they will take a submission. You have two ways past that, and the choice is yours to make once per product.
The default, and the one we suggest. The agent stops at the login wall, you sign in yourself, and it carries on with the form. Nothing exists in your name that you did not watch being made.
A pause on every platform, which is the slow part of a twenty-platform run.
Tell your agent which Google address to use for this product. Where a platform offers "Continue with Google" it clicks it and picks that account, and on most of them that same click is the sign-up. It asks once, at the first submission, and carries your answer through the rest of the run.
Every listing it creates is tied to that address for good, so name the one you would want on a company profile. Any other login, X, Apple, or an email and password, still comes back to you: the agent will not type a password.
Either way, none of it comes near us. It is your agent, in your browser, on your machine: the account you name is spoken to it, not stored here, and SubmitMap never holds a password, a session or any access to the accounts your listings end up under. What reaches us is the record you asked it to write down. The rest of that line is here.
Start it where the product lives
You can run all of this from anywhere. Running it in the product's own repository is worth three things you would otherwise do yourself.
Started in your repository, an agent has the README, the package metadata, the site copy and the assets folder in front of it. Most of what a submission form asks for is already written somewhere in there, so it drafts from that and asks you only about the rest.
A submission form takes a file, and a file input cannot be handed a URL. An agent that can see your disk uploads the logo and the screenshot from where they already live, which is the step that otherwise ends with you downloading your own images.
Platforms that want their badge or a link on your site before they will approve you are the ones people give up on, because the work is on the other side: open the editor, find the footer, add the markup, deploy, then go back and click verify. With the repository open the agent makes that change, shows you the diff, and you commit it. Or you tell it to do the lot and check afterwards.
The badge one is not a rare case: 87 of the 242 platforms in the directory want a badge or a link back on your site, several of them before they will even look at your submission. There is a prompt for it in the guides.
Have these ready
- A live public URL. Almost everything in the directory refuses a product it cannot open.
- A square logo and a wide screenshot, as files on your machine, not only as images on your site.
- An inbox you can actually reach: confirmations, verification links and rejections all arrive by email.
- A decision about logins: you sign in at every platform, or you name a Google account and let the agent do it.
- A decision about your site: whether the agent may edit it to add a badge, or should hand you the change to make.
Missing one of these is not a blocker, it is a question your agent will stop and ask you halfway through a form. The submission pack page is the longer version of that list, with the sizes and the character limits.
NextSetup