Privacy policy
This page covers the account holder’s own data. Visitors of a client site are never tracked by Plugster: no cookie, no identifier, no profile is built about them, and no identifier of theirs is written to the database. When they question a site through an AI assistant, the call journal keeps the question as it arrives (a search, a product asked about), with no identifier that ties it to a person. A caller’s address, or the network it belongs to, is held in memory for an hour at most, to cap how often the same caller may call, and never stored by the service itself. The web server in front of it (nginx) does keep its access logs, which carry the address of every request, including a visitor’s browser loading the script of a client site: they are kept 14 days for security only, neither analysed nor tied to an account or a visitor, and deleted past that.
Data controller
TooLooLoo, Raphaël Hue, 29 rue des Trois Rois, 44000 Nantes, France. Contact: hello@plugster.ai.
Data collected
Account email and, once one is chosen, a hashed password; for the hour it stays valid, the sign-up link asked for from the home page, with the site address typed, the language and the promotion code; for the hour they stay valid, the links sent to the account’s address to sign in to it, delete it or change its address (with the new address asked for); the language and the promotion code the account signed up with, and whether it signed up through the home page chat; the Stripe customer and subscription identifiers and the subscription’s status; the sites declared and their crawl results; a journal of the tool calls made against a site (date, tool, transport, agent family rather than its raw header, parameters as received, result count, latency), never the full content returned; for a site’s chat, the model key its owner gave, kept encrypted and never handed back, neither on screen nor in the export, with its provider, its dates and the state of its latest tries; the chat’s settings (on or off, model, ceiling of messages a day, look of the window) and the number of messages answered each day, and each day the products shown and opened from the chat, the add-to-cart offers made and confirmed.
Purposes and legal bases
Each processing has its legal basis (GDPR, article 6). Performance of the contract: running the account, the crawl, the tools and the dashboard, showing an owner what AI assistants search for on their site, keeping the latest checks of a site’s ownership (date, method, what they found) to help its owner complete it, and sending the email the account needs (sign-up or sign-in link, password reset or choice, confirmation of an address or of a deletion, end of a discount, suspension of a site and the reminder before its deletion, loss of ownership of a site, a reminder when a site is still not verified a few days after it was added, the instructions to verify a site when its owner asks for them). Legal obligation: billing and keeping invoices, held by Stripe. Legitimate interest: the security of the service (the web server’s logs, the caps on call frequency) and the cookieless audience measurement described below.
Cookies
Five cookies, all exempt from consent: the session cookie that keeps an account signed in (thirty days), the device cookie that remembers this browser has signed in to the account before, so that it is not locked out when someone else tries passwords on the same email address (six months), the language cookie set by the language switch (one year), and two about promotion codes (thirty days), one holding the code a campaign link carried so that it can be applied at payment, the way a basket holds what was put in it, the other remembering that the banner announcing it was closed. None is set by a third party, none is used for tracking or advertising, and none serves to build a profile.
Audience measurement
The pages of the service are counted with Umami, an open-source tool hosted on the same server as the service, not by a third party. It sets no cookie and does not store the visitor’s IP address. A page view is recorded with the page (without the parameters of its address), its title, the referring site, the browser, the operating system, the device type, the screen size, the language, and the country, region and city derived from the IP address. To count visitors, Umami combines the IP address and the browser with a salt renewed every month into a hash: page views from the same connection within a month are linked to one another, never to a person, and nothing links them from one month to the next. This measurement is exempt from consent. Nothing is shared or sold.
A client site’s chat
When a site’s owner switches it on, a chat answers their visitors on their pages. The text a visitor types goes through the service to the model provider the owner chose (Anthropic, OpenAI, Google, Mistral, DeepSeek, Groq or OpenRouter), with that owner’s key and under the terms they accepted with that provider. This text is never written to the database, to a log or to an export. What stays of a question is a short rephrasing written by the model without personal data, a theme, whether the answer came from the site, and the pages or products cited; a filter of the service also masks emails, even spelt out in words, domain names and links, postal addresses, phone numbers and runs of digits. If the visitor says whether an answer helped them, that feedback is kept with the rephrasing, with no identifier but the answer’s own, drawn at random. Nothing ties two questions of the same visitor together. The detail is erased after 30 days; only counts by theme and by day remain. The service also counts, by day and with no identifier, the product pages opened from the chat and the add-to-cart confirmed. For their visitors, the site’s owner is the controller and Plugster its processor. Limit: the rephrasing is written by a model, which a visitor can push into repeating their own words, and the filter catches the common forms of an email, an address or a number, not a name.
The home page’s AI test
On the home page, anyone can ask what an AI says about a site, without an account. The site’s address, the trade and the city typed go to Anthropic (Anthropic PBC, United States), with Plugster’s key. Anthropic uses them to search the web and open the site, and the answer is shown to whoever asked. The service writes them neither to the database nor to a log. It keeps the question and its answer in memory for three days at most, to answer whoever asks it again without a new call; a restart of the service erases them sooner. The IP address of whoever asks is kept in memory for one hour, to limit the questions from one address to five, and is never stored. Only counts by day remain, with no identifier: the questions asked, the sites cited or not, the refusals, the moves on to the sign-up and the cost of the calls. Anthropic may process this data outside the European Union, under the standard contractual clauses its terms carry. Legal basis: Plugster’s legitimate interest in showing what its service does.
Processors
Four providers see part of an account holder’s data. OVHcloud (OVH SAS, France) hosts the server, the database and its backups, in the European Union. Stripe (Stripe Payments Europe, Ireland) sells the subscription in its own name, as reseller of the service, and handles payment, invoicing and VAT: it receives the account email and its billing details, as an independent controller for those, under its own privacy policy; Plugster never sees or stores a card number. Resend (Resend Inc., United States) sends the transactional email and receives the recipient address and the message. Anthropic (Anthropic PBC, United States) answers when the account holder asks for help pasting the lines in the home page’s conversation: it receives their question, their site’s address, what the check finds there and the start of their home page’s header, with Plugster’s key; the answer is shown to the account holder, and the service keeps neither the question nor the answer. Stripe, Resend and Anthropic may process data outside the European Union, under the standard contractual clauses their terms carry. Beyond these four, only the model provider a site’s owner chose receives data, and only while that site’s chat is on: what the section on the chat describes, with that owner’s key and under the contract they made with that provider, which is therefore not a processor of Plugster. Nothing is ever sold or used for advertising.
Retention
The web server’s access logs and the system log are kept 14 days. The tool-call journal is kept 90 days and purged past that age. A sign-up, sign-in, deletion or address-change link is erased as soon as it is used, otherwise after one hour. Rephrasings of the questions asked to a site’s chat, with the visitor’s feedback on the answer, are kept 30 days, then erased. Counts by provider and by model, which name no site and no account, remain after a deletion to measure the quality of the service’s answers. Account and site data are kept for as long as the account exists; those of a site suspended because the account’s plan no longer covers as many, and that its holder has not kept, are purged 60 days after the suspension. All are purged immediately, with no grace period, when the account is deleted, the subscription being cancelled at the same moment. Two copies outlive that purge for a while: the database backups, kept 14 days on the server and 60 days in an encrypted copy off the server, after which the deleted account is gone from them too; and the customer Stripe holds (email, invoices), which Stripe keeps for as long as its own legal obligations require, ten years for invoices.
Rights
An account holder can access, correct or delete their data from the profile page at any time, and download it as a JSON file (account, sites, settings, calls counted per day) (portability). They can also object to a processing based on legitimate interest, or ask for one to be restricted. For anything the page does not cover, write to hello@plugster.ai; a complaint can also be lodged with the CNIL (France’s data protection authority).
Contact
hello@plugster.ai