cMail docs

Privacy and data-handling guide

Last reviewed: 14 August 2026

This guide helps a deployment operator prepare its own privacy notice, records of processing, provider review, and individual-rights procedure. It is not a ready-made legal notice or legal advice. The organisation must adapt it to its jurisdiction, contracts, workforce, users, message purposes, and actual Cloudflare/identity/outbound-provider configuration.

cmail is self-hosted software. The upstream project does not receive a deployment's accounts or mail merely because the software is used. The deployment operator normally determines why and how personal information is processed. It must identify its own legal role. Cloudflare, Google, Microsoft, Postmark, support providers, DNS/reporting services, and backup destinations may have separate roles under their agreements and the operator's law.

Data inventory

Review the deployed schema, provider settings, logs, and local modifications. A typical cmail deployment can process:

Category Examples Typical purpose
Account and identity Name, sign-in email, provider and immutable provider subject or bound email identity, hashed one-time codes, account status, role, sessions Enrolment, authentication, authorization, support, offboarding
Organisation directory Position title, work email, layer, unit, role, reporting structure, public/internal setting Internal administration and an optional public directory
Mail content Sender/recipient fields (including Bcc, kept envelope-only), subject, plain text, sanitized HTML, attachments, inline images, drafts, signatures Receiving, composing, storing, finding, and delivering organisational mail
Mail-flow and security metadata Envelope addresses, message identifiers, size, provider response, authentication results, spam score, suspicious-link/risk information Delivery, investigation, abuse prevention, troubleshooting
Administrative evidence Manager action, actor, target, timestamp, source address, policy version and acceptance, travel-approval decisions Security review, accountability, policy administration
AI app connections (MCP), when enabled App name and identifier, approved scopes and mailboxes, hashed tokens and codes, connection and last-use times, a per-call log of tool, mailbox and outcome (no content) Letting a person's chosen AI assistant work with their mail, revocation, accountability
Device notification data Push endpoint, keys, browser/device capability and subscription timestamps Optional new-mail notification delivery
Operational configuration Organisation name, domains, support contacts, provider choices, approved sign-in countries, retention and quota settings Running and supporting the deployment

Do not collect extra data merely because a field or provider feature is available. Do not copy secrets, raw access tokens, message bodies, attachments, or push endpoints into tickets or audit packs unless strictly required and approved.

Storage and disclosures

This section lists where cmail stores data and who else can access or receive it.

Record the actual regions, subprocessors, transfers, and contracts for the deployment. Source documentation cannot establish those facts.

Public directory privacy

The public directory is off unless the global setting is enabled. Positions are internal by default and must be explicitly made public. A public position may expose only the approved occupant name, work email, and position title. Account identifiers, sign-in addresses, permissions, reporting details, and other personal information remain internal.

Before enabling publication:

Offboarding makes the person's public positions internal. A manager must still review successor details and any separately published copies.

Retention, deletion, and preservation

The operator chooses separate periods for deleted mail, attachments, mail trace, and application audit records. Scheduled retention is off by default. Do not enable destructive cleanup until periods, backup, recovery, privacy, records, litigation/preservation, and contractual requirements are approved.

cmail has no legal-hold workflow and its application audit log is not an independently immutable archive. If preservation, eDiscovery, WORM (write-once, read-many) storage, or protected export is required, implement and test it outside cmail before destructive retention is enabled.

AI app connection records are removed 30 days after they are revoked or expire, expired tokens after a day, and the per-call tool log after one year.

Offboarding revokes access and disables the personal mailbox. It does not automatically erase retained mail or every backup. Document the difference between access removal, mailbox disablement, ordinary retention, backup expiry, and an approved deletion request.

Individual requests and organisational mail

Publish a contact and verified procedure for access, correction, objection, restriction, portability, or deletion requests that apply in the operator's jurisdiction. A request may involve other people, confidential correspondence, records duties, legal privilege, or security evidence. Do not provide raw database/R2 exports without review and redaction.

The operator should be able to locate a person's account, owned mailbox, policy acknowledgements, public positions, mailbox grants, and relevant audit records. Mail content and backups require a separately controlled search and review process. cmail does not provide a complete privacy-request or eDiscovery workflow.

Operator notice checklist

An organisation-specific notice or internal policy should accurately state:

Retain the approved notice with the deployed commit and the evidence checklist in Security, privacy, and assurance. Re-review it whenever the domain, provider, data location, purpose, public directory, retention, monitoring, backup, or support arrangement changes.

← Documentation home