Skip to content

Security

Your conversations, handled with the care you'd expect.

A plain account of what Senvoro accesses, where it goes, who can see it and how to remove it. Including what we haven't done yet.

  1. 01

    You connect

    On Google's, Microsoft's or Slack's own permission screen. We never see your password.

  2. 02

    Tokens are encrypted

    Access tokens are encrypted with AES-256-GCM before storage and can't be read by any browser, including yours.

  3. 03

    Messages sync

    Into a database where row-level security limits every row to its owner.

  4. 04

    AI reads, with limits

    Text is sent to Anthropic for prioritising, linking and drafting. It isn't used for training and the model can't take actions.

  5. 05

    You decide

    Nothing is sent, archived or deleted at the provider unless you do it.

What we access

Only the accounts you connect, with the narrowest permissions each platform offers that still let Senvoro read your conversations and send replies you approve. For Gmail that's read and send (not delete or manage settings); for Outlook, Mail.Read and Mail.Send; for Slack, history for conversations you're in and posting as you. The full list for each platform is on the integrations page.

We sync the last 30 days on first connection, then keep up with new messages. We don't download attachments, only their names.

Encryption

All traffic uses TLS. Provider access and refresh tokens are encrypted by Senvoro with AES-256-GCM, using keys held outside the database, before they're stored. Database permissions stop them being returned to any browser, including yours. Our database host encrypts stored data at rest.

AI processing

Senvoro uses Claude, by Anthropic, to prioritise conversations, connect related ones, extract next steps, draft replies and answer questions. Message text is sent to Anthropic's API for this. Under Anthropic's commercial terms it isn't used to train models, and we never train models on your messages ourselves.

Messages you receive are treated as untrusted: they're passed to the model as data, never as instructions, and the model has no ability to send, fetch or change anything. Every citation it returns is checked against the messages it was given, and a deadline is only shown when its exact words appear in the source.

Sending and approval

Senvoro never sends anything to another person on its own. Every outgoing message requires you to press “Approve and send”, and the approval is recorded in your workspace's activity log before the message is handed to the provider. There is no auto-send setting.

Who can see your data

Your workspace is isolated at the database level: every row carries its owner, and row-level security rejects reads from anyone else. Our policy is that staff don't access message content. If we ever need to for support, we'll ask you first.

Disconnecting and deletion

Disconnect an account in Settings and we revoke the grant at the provider where it allows (Google and Slack do; for Microsoft, remove the app at myaccount.microsoft.com), delete the stored tokens, and delete every conversation, message, draft and action that came from it.

You choose how long synced messages are kept (30 to 365 days, 90 by default). Older messages are deleted daily. You can export everything as JSON, or delete your whole workspace in one step. Deleted data can remain in encrypted database backups until they expire; we'll publish that exact window before general availability.

Certifications

Senvoro is a young product. We don't hold SOC 2, ISO 27001 or similar certifications yet, and we haven't had a third-party penetration test. Before Gmail can be connected by more than a limited number of users, Google requires an independent security assessment (CASA) of apps that read mail. We'll list each assessment here when it's complete, not before.

Sub-processors

ProviderPurpose
VercelApplication hosting
SupabaseDatabase and sign-in
AnthropicAI processing (Claude)
Google, Microsoft, SlackOnly the services you connect

Reporting an issue

Found a vulnerability? Contact us and we'll respond quickly. We'll acknowledge within two working days and won't take action against good-faith research.

Read our privacy notice for legal bases, retention and your rights.