Replaces Exchange Server, IMAP and POP
ChainMail
A mail server your Outlook already knows how to talk to.
ChainMail is a multi-tenant mail platform that speaks the protocols Outlook expects — MAPI over HTTP, Autodiscover, EWS — alongside plain IMAP and POP3. Mailboxes replicate between machines continuously and fail over to a standby without anyone re-entering a password.
Running in production across eight domains and about 126 mailboxes. Classic Outlook connects to it natively, verified against a real client.
- Protocols
- MAPI over HTTP, Autodiscover, EWS, IMAP4, POP3, ActiveSync, SMTP
- Clients
- Classic Outlook with no plugin, Rivet Mail, any IMAP or POP client
- Replication
- ZFS snapshots streamed continuously, one dataset per tenant
- Failover
- Pacemaker with hardware fencing, proven in three live drills
- Deployment
- Single server, replicated pair, or hosted by us
- Tenancy
- Multi-tenant, per-tenant branding, reseller hierarchy
Outlook connects without a plugin
ChainMail answers Autodiscover and serves MAPI over HTTP, which is what Outlook looks for. Users type an address and a password; nothing is installed, and no compatibility layer sits in between.
Every mailbox is replicated as you use it
Each tenant gets its own filesystem dataset, snapshotted and streamed to a second machine continuously. A failure loses the seconds since the last snapshot, not the day.
Failover that has actually been tested
The standby takes over under Pacemaker control, and the failed node is fenced at the hypervisor so it cannot come back and write to the same storage. We have run it three times against real machines rather than assuming it works.
Shared mailboxes and delegation
Shared inboxes, calendar sharing, send-as and delegate access work the way people expect coming from Exchange, because the protocol underneath is the same one.
An admin panel, not a config file
Domains, mailboxes, aliases, forwarding, quotas and distribution lists are managed from a web panel. Resellers get their own scoped view of only their own customers.
Run it on your own hardware
The same build runs on a single server in your office, on a replicated pair, or on our infrastructure. Moving between them is a migration, not a rewrite.
Listed so you can decide with the whole picture rather than discover it after signing.
- Hypervisor-level high availability — failover is currently proven between two virtual machines that share one physical host
- Per-mailbox calendar permissions tightened further