Your data stays yours,
and every action is on the record.

How Humonix handles your data, how you sign in, what agents can and cannot do, where things are hosted, and what we have not done yet.

Exportable
Your Brain, its sources, and every log, in open formats, any time you ask. When you leave, they come with you.
Not used for training
Your data is never used to train shared models.
Least-privilege access
Each module gets only the access its job needs. The ads agent can change budgets within a cap; it cannot read your accounting.
Approval rules you set
Anything that spends money, changes a price, or goes public waits for a named person.
Audit log
Every action and approval, with what it was based on, who approved it, and when.
Prompt-injection handling
Text inside emails, web pages, and documents is treated as data, never as instructions to an agent.

Your client account

  • No passwords to steal: you sign in with a 6-digit code sent to your email. Codes are stored only as a keyed hash, expire after 10 minutes, allow 5 tries, and 10 failed tries in a day lock that email out.
  • Sessions live on our server, not in your browser: signing out, or an owner removing someone, ends their access at once. The cookie is HttpOnly and sent only over HTTPS; sessions last 14 days.
  • Nobody gets in uninvited. We invite your first owner; owners invite the rest. One email belongs to one company, and every page reads your company from your session, never from a link.
  • Your access log shows who opened or downloaded what, and when.

When agents work for you

  • Agents draft; people decide. Replies, reminders, nurture emails and repairs wait in an approval queue, and the approval is logged.
  • We never send email in your name. Nurture emails are drafted for your person to send from their own mailbox, and an owner on your side approves the exact words first; changing a word clears the approval.
  • Every Outbound module nurture email carries the sender, your postal address and an unsubscribe link. An unsubscribe takes effect at once and withdraws anything still waiting for approval.
  • An Outbound module contact gets emails only with a recorded consent basis and date (Canada’s anti-spam law, CASL, applied in the US too). Click tracking is off unless you turn it on. Sending from the Outbound and Grow modules stays off until the same footer and consent checks are switched on for your account.
  • Each client’s records are kept apart: every table is keyed to your account, and your portal never shows another company’s activity.

Where things run

  • Our own systems, including the lead intake and customer records behind this site, run on Fly.io in Toronto, Canada, on encrypted volumes with continuous backups and daily snapshots.
  • Hosted close to home, in North America: your Brain and client data are hosted in Toronto by default, like our own systems. US hosting for US clients is on request: a client’s own instance in Virginia (US) holds its synced records, files and search index, while the app, sign-in, customer records and the ledger stay in Toronto. Canadian clients’ data is hosted in Canada (Québec Law 25). AI models run with US providers (Anthropic first), listed on the subprocessors page.
  • AI models are called through their providers’ APIs under business terms that exclude training on your data.

Law 25 and CASL

  • We collect only what a form asks for, and marketing consent is recorded separately from the request itself, with its wording and time.
  • Every marketing email and text includes identification and an unsubscribe, as Canada’s anti-spam law requires.
  • You can ask us at any time to export or delete the personal information we hold about you.

Our crawler

  • Our crawler reads public homepages, about once a month. It names itself in every request and links to a page explaining what it does.
  • It obeys robots.txt and never works around a block, a login or a rate limit; a site that turns it away is left alone.
  • It keeps what it found (which tools a site uses, a few scores), not copies of the pages.

What we have not done yet

  • SOC 2: not certified yet. We build toward SOC 2 controls; ask us for the current status.
  • No third-party penetration test report to share yet.
  • No paid bug bounty program.

Report a vulnerability

Found a security issue? Tell us with the button below (it opens our contact form marked security); we typically reply within one business day. Please don’t access or change other people’s data, don’t slow the service down, and give us a reasonable time to fix it before you publish. Our security.txt is at /.well-known/security.txt.