Skip to content
BI Beeby HYVE Labs

Security

Know where your data goes.

Every workspace is private, every role is checked on the server, and the AI model never sees raw rows. This page lists what is implemented today, as it works.

No certification such as SOC 2 or ISO 27001 is claimed. Owners and admins can export their workspace's security evidence from the app.

Your sources

Databases, CRM, sheets, filesPostgres, BigQuery, HubSpot, Google Sheets, S3, CSV
Encrypted loginsAES-256-GCM, never sent to the browser

Your BI Bee workspace

Datasets and dashboardsChecked columns, personal ones marked
PeopleFour roles, sessions, activity log
AI toolsRead-only tokens, every call logged
External AI model · on by defaultField names, summary figures, a few category values and the question. Never raw rows.
Hosted on Vercel with a Supabase PostgreSQL database. Every query is limited to your workspace. The AI model is on by default for new workspaces, and an owner or admin can switch it off.

Controls

Twelve controls, all live today.

  1. 01

    Private workspace

    Every source, dataset, dashboard, report and audit query needs the signed-in workspace. One customer's data never appears in another's.

  2. 02

    Encrypted source logins

    Passwords, API keys and OAuth tokens for your sources are encrypted with AES-256-GCM on the server and never returned to the browser.

  3. 03

    Four roles, checked on the server

    Owner, admin, analyst and viewer. Every API route checks the role, so a hidden button is never the only thing standing in the way.

  4. 04

    Personal columns marked for you

    Email, phone, name and address columns are marked personal on import. Viewers see neither their values nor their names, on dashboards, in reports or in exports.

  5. 05

    AI on your terms

    The external AI model is on by default for new workspaces, and owners and admins can switch it off. It receives field names and types, summary figures such as totals, ranges and counts, up to eight common values of open category fields, and the question; never raw rows or values of fields marked personal or confidential. Any figure in an analyst answer must match your data.

  6. 06

    Revocable AI tool tokens

    Claude, Cursor and other MCP clients connect with read-only personal tokens. Owners and admins see every token in the workspace and can revoke any of them.

  7. 07

    Activity log

    Sign-ins and failed attempts, sources, syncs, dashboard edits, reports, team and billing changes and every AI tool call are logged. Owners and admins can export the evidence.

  8. 08

    Rate limits

    Sign-in, sign-up, password reset, booking and API requests are rate limited to slow down password guessing and abuse.

  9. 09

    Hashed passwords and sessions

    Passwords are hashed with scrypt and only hashes of session tokens are stored. You can see your active sessions and sign out other devices.

  10. 10

    Old data deleted on schedule

    Each sync keeps earlier snapshots for 7 days so you can roll back, then deletes them. Deleting the workspace deletes all of its data.

  11. 11

    Safe spreadsheet exports

    CSV exports neutralise spreadsheet formulas, so a cell cannot run a formula when the file is opened in Excel or Sheets.

  12. 12

    Hardened web layer

    Strict security headers (CSP, HSTS, no framing), same-origin checks on every change and signed webhooks from connected stores.

In practice

Who sees what. And a record of who did what.

Pick a role to see what it opens, then see the kind of entries the activity log keeps.

Roles · one workspace1 of 6 areas open
  • DashboardsView, filter, export
  • Data sourcesNo access
  • AI analystNot available
  • Personal columnsHidden, names too
  • Team and rolesNo access
  • BillingNo access
Full Limited Closed
Activity logExample rows
TimeWhoActionResult
09:02Ownerauth.loginSigned in
09:05Analystconnector.createPostgres · logins encrypted
09:11Admindataset.governance.updatecustomer_email → personal
09:24Unknownauth.login.failedWrong password
09:40Analystmcp.tool_callquery_dataset · totals only
10:02Analystdashboard.undoRestored the previous version
Recent activity shows in the app's Trust center, and owners and admins can export it as evidence.

Data lifecycle

From sign‑up to deletion.

  1. Step 1

    Sign up or join

    An owner creates the workspace. Everyone else joins by invitation, with a role.

  2. Step 2

    Connect

    Source logins are encrypted, and each sync is recorded with its row count.

  3. Step 3

    Classify

    Columns are typed and personal ones marked. Mark any other column confidential.

  4. Step 4

    Work

    Dashboards, reports, exports, the analyst and AI tools follow each person's role.

  5. Step 5

    Remove or delete

    Removed members lose access on their next request. Deleting the workspace deletes its data.

Bring your security questions.

We will walk your IT and data team through these controls on a call, or answer by email.