agenticboxes.email How it works Pricing Compare Quickstart API Reference Audit Security Changelog Get an API key

Security & data handling

Last reviewed 16 October 2026. Everything below describes what is running today, not what is planned. Where we have not done something, it says so.

What we store

WhatWhereEncrypted
Message bodies and attachmentsAmazon S3Yes — AES-256 at rest, and inbound mail is additionally encrypted with a KMS key before it is written
Message metadata (addresses, subject, timestamps, message ids)Amazon Aurora PostgreSQL 16Yes — storage encryption enabled on the cluster
Your API keysAuroraStored hashed. We cannot read your key back to you; a lost key is rotated, not recovered
Card detailsNot stored by us at allStripe holds the card. We hold a token and the caps you set

What deleting actually removes

This is the part most pages are vague about, so here it is precisely.

  • Delete a message (DELETE /api/v1/messages/:id) removes the stored body from S3 immediately and clears the pointer to it.
  • Delete with the purge flag ({"purge": true}) also clears the subject and strips the display names from the From, To and Cc addresses.
  • What remains either way is the audit record: the message ids, the timestamp and the direction. That is enough to prove a message existed and when, without disclosing what it said or who it named. It is deliberate — it is what makes the audit trail worth anything.
  • Raw inbound mail is held in a separate bucket for one day and then deleted automatically. That copy exists only so a processing failure can be replayed.
  • Close your account and the account data goes with it.

Who else touches your data

ProcessorWhat for
Amazon Web ServicesAll hosting, storage and mail delivery. SES sends and receives, S3 stores bodies, Aurora stores metadata, KMS holds the encryption keys. Everything runs in us-east-1.
StripePayments, including the agent wallet. Card details go to Stripe and never to us.
HeyCatch (PostHog)Website analytics on these pages only — which pages are viewed and what is clicked. It never sees your mail, your API keys or your account data. We run it cookieless: it sets no cookie and stores nothing on your device beyond the current tab session, so there is nothing to follow you between visits.

We do not sell data, and we do not share it with anyone outside that table.

Analytics note, 16 October 2026: the analytics SDK ships with a first-party cookie carrying a persistent device id. We turned that off deliberately (persistence: 'sessionStorage'). The trade is real and worth stating: we can see how a visit moves through the site, and we cannot tell a returning visitor from a new one. We would rather have the weaker number.

Sending safety

  • Permission-based mail only.
  • Every account carries its own sending reputation and its own suppression list. One account's bounces or complaints never affect another's.
  • An address that bounces or complains is suppressed for that account automatically, so it is not tried again.
  • Abuse is isolated and suspended rather than spread across the platform.
  • DKIM, SPF and DMARC are configured for your domain as part of provisioning.

The audit trail

Every account's platform events are written into a hash chain, so an event cannot be altered after the fact without breaking the chain. A reconciliation job re-verifies every chain on a schedule and publishes the result, including failures.

That score is public and machine-readable, and it is the same file this site reads: the audit page · audit.json. When a run has flagged something, the page has said so — there is a worked example of exactly that in the published history.

Compliance: where we actually are

Plainly, because the honest answer is more useful than a badge:

ItemStatus
SOC 2Not held, and no audit is in progress. If you need a SOC 2 report from your vendors today, we are not there yet.
ISO 27001Not held.
HIPAA / BAANot offered. Do not send protected health information through this platform.
GDPR-style deletionSupported today. The purge behaviour above is the mechanism, and it is in the API rather than a support request.
Encryption at restIn place on both the database and the object storage, verified 16 October 2026.
Encryption in transitTLS on the API. Mail uses opportunistic TLS, which is what SMTP allows — no provider can promise more than the receiving server supports.
Tamper-evident audit logIn place, and published.
Penetration testNo third-party test has been commissioned.

The underlying platform has run on AWS since February 2021 and has handled hundreds of thousands of messages. That is operating history, not a certification, and we are not going to present it as one.

Reporting a problem

Email support@agenticboxes.email, or use the machine-readable disclosure address at /.well-known/security.txt. Agents can file a ticket through the API: POST /api/v1/support/questions.

Related: Privacy · Terms · Refunds

Product

  • How it works
  • Pricing
  • Changelog
  • Audit & integrity

Developers

  • Quickstart
  • API Reference
  • Overview
  • OpenAPI spec
  • Agent skills

Company

  • Receipts
  • AgenticBrian.com
  • Brian on LinkedIn
  • Terms · Privacy · Security

Contact

  • support@agenticboxes.email
  • Reaches an agent on this platform, and a human behind it.
© 2026 AgenticBrian Holdings · Built and run by Brian Becker