Skip to content
Site language : English, EN
Subscribe to Solo

Security of your data

Updated: November 20, 2026

This page is translated from French; the French version is authoritative. Français

Your agents read your emails, your files and your accounts. Here is where they work, who can see what, and what happens the day you leave. This page restates the written commitments of our terms and conditions, our data processing agreement and our privacy policy, which prevail, and explains how we keep them today.

In France, with OVHcloud

  • Hosted version: your control center and your agents' machine are hosted in France, on a dedicated OVHcloud server in Roubaix. Your agents have a virtual machine of their own there, with their browser and their working files.
  • Installed version: if you have your own server, your agents' machine and your projects' machine run on it, on a dedicated OVHcloud server rented in your name, in France; your control center stays with us, in Roubaix.
  • Backups: kept off the hosting server, also with OVHcloud, in its Strasbourg data center.
  • Your customer account and your messages to our team: your Lastnod subscription and what you write to our team (tickets, tasks you assign in your space to the role declared for us, emails to team@lastnod.com) are handled by our own installation, on another OVHcloud server, in Gravelines. The end-of-service export is also prepared there (below).
  • For the data in your space, OVHcloud is our only processor. We tell you before any change.
  • What your agents read in order to work also goes to your AI provider, outside France: see "What the AI reads" below.

Your space, separate from all others

  • One control center per customer, under its own system account, with its own database, processes, secrets and accounts. No other customer can access it, except what you share with them yourself (a chat room, a project opened to a collaborator).
  • In the hosted version, your agents work in a virtual machine that is yours alone. Two customers' machines cannot see each other on the network. Your agents' machine reaches no internal machine of Lastnod or of another customer: only the internet, like an office computer, and your control center, over HTTPS.
  • Each agent identifies itself with a token specific to your space: the server derives the company and the role from it, never from a text. Access by one customer to another's space is refused, and our tests check it.
  • Every human account is protected by a second factor.
  • The credentials of the accounts you open to your agents are stored encrypted in your space's vault: they are never displayed and never go through a conversation or an email. You connect your AI subscription yourself, with its provider: its credentials never go through us.
  • You choose what your agents do on their own and what you approve yourself. Whatever you reserve for your approval (sending, publishing, paying) is refused by the server until you have approved it.
  • Every action in your control center is recorded in a log (who, when, what) that you can consult in your space; our administration commands run directly on the server go to the server's audit log instead.

Backups: every night, encrypted, tested

  • Every night, your data is encrypted on the server, with a key specific to your installation that we keep off the hosting server, then copied to OVHcloud object storage in Strasbourg. What is backed up: your control center (its database, files and library) and, in the hosted version, your agents' machine (their folders, and their browser profile with the accounts you have signed into). Only the connection to your AI subscription (Claude or ChatGPT) is not included: after a restore, you sign in again.
  • Each copy is locked there for thirty days in compliance mode: no one can delete or modify it before then, not even us. It is deleted afterwards, at the latest thirty-one days after it was sent.
  • A backup is only worth something if it can be restored. Every week, we restore the latest control center backup of one installation, in turn, into a disposable database separate from production: we check its fingerprints, compare its number of tables and rows with production, check that its file archives can be read in full, then delete the test copy. This check is automatic and keeps only numbers; a failure alerts us.
  • In the installed version, your control center is backed up the same way. In addition, every night we copy your agents' machine and your projects' machine in full, encrypted, to the backup space provided with your server, which keeps the copies of the last seven nights, one per week for five weeks and one per month for three months.

What we see, and what we don't look at

We run the server: like any hosting provider, we can technically access it. Here is how that access is limited.

  • No permanent account in your space. For troubleshooting, you open a support access yourself, confirmed by your second factor, with a written reason and the duration you choose, twenty-four hours at most, read-only or over your whole space. You revoke it whenever you like, and every action taken through it appears in a log you can consult. Even when open, it cannot change your accounts or second factors, start your agents or change what they do on their own, or see your secrets. Only Eric Mollenthiel uses it, from his own browser: our agents may ask you for it, never use it.
  • Operations: our deployment console keeps technical access to the service to keep it running (updates, backups, repairs). It is not used to read your messages, your files or your calendar. We only read your data through the support access you open, during a service you order from us (setting up the installed version, for example), only as far as it requires, or, where necessary, for the security of the service, to examine a reported abuse (article 10 d of the terms) or to comply with the law.
  • Who has this access: only Eric Mollenthiel and the agents of the Lastnod team, who work from his server, have technical access; no other person on our side. These agents keep administrator rights on the servers and the keys to the backups: what stops them from reading your data is a written rule that binds them. They open neither your databases, nor your files, nor your messages, nor your agents' browser, nor the logs that carry your data; they work on status, settings and numbers, and their tools display nothing else. So they read no personal data from your space or your server, except the tickets and requests you send us and the report your installation sends us (below). A repair that would require reading your data is done by Eric Mollenthiel himself, or by an agent of your space, with your AI subscription. What they read while administering the servers goes to the providers of their AI subscriptions: see "What the AI reads".
  • Our servers are administered by SSH key only, passwords refused, from a single address, with the root account closed to remote login, behind a firewall, with automatic security updates; every command run as administrator is recorded in a log, kept for twelve months. If you have your own server, we also administer its host machine and your agents' machine, and keep emergency access to your projects' machine, which your agents run; as the holder of its OVHcloud account, you keep its emergency access at OVHcloud.
  • What your installation sends us: the list of its accounts (first name, email address, date of last visit), its status, and usage numbers counted for the whole installation, per month. Never a title, a text, a file or a secret. Details are in our privacy policy, under "Your accounts and the use of your installation".

What the AI reads, and with which subscription

  • Your agents work with your own Claude subscription, from Anthropic, or ChatGPT subscription, from OpenAI, in your name, directly, with no Lastnod account or server in between: Lastnod does not resell AI. You choose each agent's AI; no other AI runs your agents.
  • What your agents read in order to work (an email, a file, a page) is sent to their AI's provider, Anthropic or OpenAI, which processes it, notably in the United States, under your contract and your settings with them, outside our data processing agreement. We advise you to refuse the use of your conversations for training their models: this setting remains yours, in the privacy settings of your Claude or ChatGPT account.
  • The HQ lead, the agent who welcomes you and organizes your Lastnod, does not review your conversations with your agents to draw lessons from them. When it helps you and notices that something is missing from their method, it may write a lesson, with no name or data from your business, never to assess anyone; it only reaches us if your company turns on lesson sharing, or if you send it to us yourself.
  • We do not use your data to train an AI model, for advertising or prospecting, or for another customer, except the lessons your company chooses to share with us.
  • Our own agents, those of the team that builds Lastnod, read your tickets and the requests you send us, your emails to team@lastnod.com, your Lastnod subscription data and your installation's report, and administer the servers, with our own AI subscriptions (currently with Anthropic and OpenAI), whose providers receive what they read as controllers of their own processing (our privacy policy, under "Who receives your data"). The use of these exchanges to train their models is turned off on every account these agents work on.

Your data stays yours: export and deletion

  • At any time, you download from your space the complete export of what belongs to you, in open formats (JSON and Markdown).
  • At the end of the service, you receive this export on request within fifteen days, through a download link valid for seven days; the archive, prepared on our Gravelines server without being read, is deleted there when the link expires. After it is delivered, or thirty days after the end if you do not request it, we delete your control center and your agents' environment, with the keys to your connected accounts. We destroy the encryption key specific to your installation, which makes your backups unreadable; they are purged when their lock expires, at the latest thirty-one days later. We certify the deletion to you in writing.
  • What remains: our servers' security logs (IP addresses, accounts and administration commands, without any content from your space), for twelve months, and contractual and accounting documents, for ten years.
  • In the installed version, we also delete your agents' machine from your server; we give you what you need to read that server's nightly copies, then we remove our access to that server. The server, your projects machine and their content remain yours.

What we don't promise

We protect your data with care, without being able to promise absolute security (article 6 of the terms). What your agents read can be read by your AI provider: choosing what you open to them remains your best protection. An account that holds little information limits what they see.

If an incident affected your data

If a security breach affected the personal data in your space, we will notify you without undue delay, and at the latest forty-eight hours after becoming aware of it, so that you can report it within the legal deadlines.

A vulnerability, a question

You think you have found a vulnerability, or you have a question about your data: write to team@lastnod.com. We usually reply within two business days.

Back to home