E-commerce

A seller rarely has one account. There is an Amazon account for the United States, an eBay account opened years ago, an Etsy shop aimed at the United Kingdom, and two Shopify admin panels you keep for clients. All of them open on the same computer, usually in the same browser.

A marketplace does not see an account when you log in. It sees a browser and a connection. In one plain browser profile those are identical for every store you run, and the overlap is the thing platforms are built to notice. Their own seller policies say plainly that linked accounts are treated as one seller.

Incognito mode does not solve this. It forgets the session at the end and presents the same machine while it lasts. A VPN changes the address and leaves the browser untouched. A virtual machine per store does work, and costs you a machine per store, plus the memory to keep three of them open at once.

ARMANOS puts each store in its own profile. A profile is a folder of browsing data plus a description of a device, and both belong to that one store. Launching two profiles at once opens two browsers that share nothing.

• Cookies and storage

  One jar in a plain browser. The session of one marketplace sits next to the session of the account beside it.

• Saved logins

  One password store holding every account you have ever typed into that browser.

• Fingerprint

  One screen, one graphics card, one font list, one timezone, offered to every site you open.

• Exit address

  One IP, usually the one your other accounts have already been seen from.

• History and open tabs

  The pages one account visits land in the same history as the next one.

The scenario is chosen once, when you create the profile. Its id in the code is shopping, and it lives in packages/shared/src/index.js, so the desktop app, the server and this website read one set of numbers rather than three.

It decides the address the profile opens on, which scripts are injected into pages, which tracker list the network layer loads, whether links are cleaned of tracking parameters, and what a safe working day looks like for a seller.

The five per-action ceilings are the part that is specific to selling. They ramp over the first week together with the page counter, so a profile you made this morning shows smaller numbers than the same profile will show next Monday.

• Start page

  https://www.amazon.com, unless the profile has a start address of its own. Your address always wins.

• Injected scripts

  fingerprint and stealth. The tracker and link work happens at the network layer, not in the page.

• Tracker list

  general: 31 domains cancelled at the request level, including analytics, session recorders and crash reporters.

• Link cleaning

  On. 40 tracking parameters are stripped from URLs, among them utm_source, fbclid, gclid, ttclid and _ga.

• Daily page ceiling

  120 at full pace, reached on day seven.

• Fingerprint template

  Not forced. You choose one of the five: macOS, Windows, Linux, Android phone, Android tablet.

The scenario carries a list of platforms. It is what the defaults were chosen against, and it is what the app shows you when you pick the scenario in the create window.

These are names, not integrations. ARMANOS does not connect to any of the seven through an API, does not read your orders, does not publish listings and does not touch your payouts. A profile is a browser with a fixed identity, and everything you do inside it you do yourself or with your own scripts.

Across the five scenarios the app names twenty platforms in total. Seven of them sit here. The other four scenarios cover social media, ad accounts, freelance marketplaces and plain private browsing.

This is the whole path for a single account, from an empty list to a profile you can work in. It takes a few minutes, and the week that follows is waiting rather than work.

Two of these steps are checks rather than settings. The proxy check and the start page exist so you find out that something is wrong before an account does.

1. 1

   Create the profile and name it after the store

   Etsy, home goods, UK tells you more in a list of fifty than Etsy 3. The name is what you will read at two in the morning.

2. 2

   Pick E-commerce as the scenario

   This is the moment the ceilings, the start page, the tracker list and the link cleaning are attached. You can change it later on the profile.

3. 3

   Pick the device template

   One of five: macOS, Windows, Linux, Android phone, Android tablet. Pick the machine that seller would plausibly work from, then leave it alone forever.

4. 4

   Attach a proxy from the right country

   A shop aimed at Germany wants a German exit. Take the proxy from the library, or paste a line in host:port:user:pass form and let the app parse it.

5. 5

   Press Check and let the country be measured

   The check goes out through the same path the launch will use, including the authentication bridge and the SSH tunnel. The country that comes back is saved on the profile.

6. 6

   Let the country set the clock and the languages

   A German exit produces Europe/Berlin and de-DE, de, en. In a country with several zones the profile keeps one of them, chosen from its own id, so it is the same on every launch and on every machine you sign in from.

7. 7

   Launch and read the start page

   The first tab is served by the app itself from 127.0.0.1 under a one-time token. It names the exit IP, the country and city, the profile, the engine and the fingerprint.

8. 8

   Warm the profile up before the first login

   See the next section, then log in and keep the day inside the ceilings.

A new profile does not get the full ceiling on day one. The ramp is a fixed curve of seven fractions, 0.2, 0.35, 0.5, 0.65, 0.8, 0.9 and 1.0, applied to both the page count and every action ceiling. Day eight and everything after it sit at the full number.

Warming up also means giving the profile something to remember. A browser created five minutes ago, with an empty cookie jar, no history and nothing cached, is unusual by itself. The cookie robot browses ordinary public sites in that profile first: 30 addresses in six categories, one of them a shopping group of eBay, Etsy, AliExpress, IKEA and Zalando. Nothing in the pool requires a login, because a robot landing on a sign-in wall produces failed sessions instead of history.

Order and selection are random per profile, so two profiles never walk the same path in the same order. The dwell on each page varies between 0.6 and 1.4 of the seconds you asked for, and each site gets six steps: open, settle, scroll, pause, scroll, pause. A fixed rhythm would be a pattern of its own.

Your accounts already live in a table somewhere. The import reads that table instead of asking you to retype it. Column names are matched by meaning, not by an exact string: Name, Название, profile name and имя профиля all reach the same field, in about twenty languages.

The import runs in two steps on purpose. The file is read and shown first: which column became which field, which columns were not recognised, a sample of the first five profiles, and the line numbers of rows that could not be read. Nothing is created until you have looked. Two hundred profiles carrying someone else's proxy are worse than two hundred profiles made by hand.

Groups arrive as names and become real groups, matched case insensitively, so Etsy and etsy do not become two folders. Proxies from the file join the proxy library, so an imported profile gets the same rotation and the same checks a hand made one gets. A one-time-code key in the file is validated and stored on the profile; a key that cannot produce codes is reported rather than saved, because wrong codes lock accounts out.

Then the list works in bulk. Select rows and you get launch, stop, move to group, assign proxy, export and delete. Bulk launch spaces the logins randomly between 3 and 14 seconds, because fifty accounts entering in one minute at even intervals reads as one script behind all of them. Only one bulk run happens at a time, and when the open-browser limit is reached the queue waits for a free slot instead of reporting fifty failures. That limit is computed from your memory: three gigabytes for the system, one per browser, clamped between 4 and 32, and you can override it by hand.

Name

Etsy, home goods, UK

Group

Etsy

Proxy

http://198.51.100.10:8080:user:password

Template

Windows

2FA

JBSWY3DPEHPK3PXP

Tags

etsy, uk

Notes

Payout account 3

The example file the app writes for you carries exactly these seven columns. Add an eighth named Scenario with the value shopping in every row, or the imported profiles land on Privacy and you set the scenario yourself afterwards.

The counter counts top-level navigations. Each profile window reports a page load to the app over an internal channel, and the app adds it to that profile's number for today. The count lives in pace.json, is written at most once a second, survives a relaunch on the same day and resets when the calendar date changes.

The bar turns amber at 80 percent of today's ceiling and red at 100 percent, and it never blocks anything. Under it you get either the warm-up day out of seven, or the age of the profile in days. That is the age of the ARMANOS profile, not of the seller account behind it, because the app has no way to know when the account was opened.

The five action ceilings are numbers on the screen, ramped by the same curve. Nothing in the app counts a listing, an edit, a buyer reply, a review request or an account switch. A browser cannot tell a page where you published a listing from a page where you read one. So those five are a target you hold yourself to, and the honest label under the bar says pages.

• Counted

  Pages opened in the profile today, per profile, resetting daily.

• Not counted

  Listings, edits, buyer replies, review requests, account switches.

• Where the numbers come from

  The scenario record in the shared package, ramped by account age. They are our recommendation, not a published platform rule.

• What happens at the ceiling

  The bar goes red. Nothing closes, nothing is refused, nothing is reported anywhere.

Accounts get linked through the payment method as easily as through a cookie. One card behind ten seller accounts is one identifier behind ten seller accounts, and no browser setting hides that. The same is true of one payout bank account, one legal name and one warehouse address.

What the app does keep apart is everything that lives inside the browser. Each profile gets its own session partition, named persist:armanos-profile plus the profile id, and on the ARMANOS Browser build it also gets its own engine directory under fork-profiles. Cookies, local storage, saved cards in the browser's own autofill and service workers belong to one profile and never reach another one. The proxy is per profile too, and WebRTC is set to disable-non-proxied-udp so a media connection does not slip out around it.

The rest is your discipline, and the app gives you a place to write it down. The Notes field and the tags travel with the profile, including through export and through sync, so which payout account belongs to which store is recorded next to the store rather than in your head. The account password and the proxy password go into the operating system keychain when it is available, and the profile file itself is written readable by its owner alone.

• Session

  persist:armanos-profile-<profile id>. Cookies, storage and browser autofill belong to one profile.

• Engine directory

  The ARMANOS Browser build runs with its own user-data-dir per profile, under fork-profiles/<profile id>.

• Exit

  One proxy per profile, measured through the same path the launch uses, with the country remembered on the profile.

• Real address

  WebRTC policy disable-non-proxied-udp, so a call does not travel around the proxy.

• Secrets

  Account and proxy passwords go into the system keychain when the system offers one. The files are owner readable only.

• The card itself

  Yours. Nothing here checks whether two profiles share a payout account, and nothing here should pretend to.