Security
in plain words.

Xorviex lets AI agents act in your tools, so what protects your account, your connections and your data should be something you can read in five minutes. This page lists what is in place today — and what is not.

Your account

Signing in takes more than a password from the first day.

  • Two-factor sign-in is on by default: a code sent to your email, or an authenticator app if you prefer, with one-time recovery codes as a fallback.
  • Your session lives in a cookie the page's own scripts cannot read, and every change you make carries a token bound to your session.
  • Changing your password asks for the current one and signs out every other browser and device. You can also sign out all other sessions from Settings at any time.
  • Repeated wrong passwords or wrong codes for one account are cut off, wherever they come from.

Your connected apps

An agent can act in an app only after you connect it, and never holds the key.

  • Tokens and API keys for Slack, Google, GitHub, Notion, Jira, Asana, Salesforce and Ayrshare are encrypted before they are stored.
  • A credential is decrypted only for the single call that needs it, on our servers. The AI model never sees it.
  • Agents are offered the tools of the apps your workspace connected, and no others. Disconnect an app and its tools are gone from the next run.
  • Google Drive access is limited to files Xorviex created or you opened with it — not your whole Drive. Xorviex cannot read your Gmail inbox.

What agents can and cannot do

Agents choose from a fixed set of tools. They do not write and run their own code.

  • Before an agent sends a message or an email, posts, commits or changes a record, a separate AI reviewer checks that exact action against your goal and your workspace rules. If it is not approved, it is not made.
  • Every agent's result is reviewed by a separate Supervisor and scored before it reaches you. Work that could not be reviewed is held for you instead of delivered.
  • Every tool call, with its arguments and its result, is recorded on the mission and can be exported.
  • You choose a mission's size before it runs, and that size is a real limit on agents, steps and time.

Keeping things apart

One workspace cannot reach another, and the open web cannot reach our internals.

  • Missions, files, memory and connections belong to one workspace. A file or a record is looked up inside the workspace that asked for it.
  • Pages are opened in a disposable, isolated browser container that is thrown away after each read.
  • Agents can only fetch public web addresses. Requests to internal or private network addresses are refused.
  • A trigger's URL is stored only in hashed form, shown once, and can be replaced at any time. Each trigger has a daily limit.

Payments and your data

We do not see your card, and your work is yours.

  • Card details are entered on Stripe's own pages. They never reach Xorviex servers.
  • Your missions, workspace memory and agent output are not used to train AI models.
  • Deleting a workspace deletes its stored files too.
  • Connections to the site are encrypted, and browsers are told to use HTTPS only.

What to know

The limits, said as plainly as the rest.

  • Text an agent reads on a web page, or receives through a trigger, can try to talk it into doing something else. Such content is marked as untrusted, and the reviewer described above never reads it — but these are safeguards, not a guarantee. Connect only the apps a workspace needs.
  • Everyone in a company workspace can read that workspace's Command Chat conversations. Invite accordingly.
  • Xorviex has not yet completed an independent security certification.

Found a problem?

If you believe you have found a security issue in Xorviex, write to privacy@xorviex.ai with “Security report” in the subject. Tell us what you found and how to reproduce it; please do not access other people's data or disrupt the service while testing. We read every report.

How personal data is handled is in the Privacy Policy.