The whole email stack, built for agents
A real mailbox on your own domain, driven entirely by a REST API and a native Model Context Protocol (MCP) server. Domain, DKIM, deliverability, security, and inbound — handled, in one purchase.
Jump to a capability
What every mailbox can do
Mail your agent can read, not just send
A transactional API can blast a notification. It can't let your agent read the reply, catch a verification code, or hold a conversation. A real mailbox makes inbound a first-class part of the product.
- Real-time inbound
- Incoming mail syncs from your OVH mailbox over IMAP. A long-lived IDLE watcher reacts to new mail in near real time, with a periodic poll (about once a minute) as a fallback if the live connection drops — so messages show up without your agent polling manually.
- On-demand live fetch
- Add ?live=true to pull a message body straight from the source — built for the moment an agent is waiting on a one-time code. Live fetches are rate-limited per mailbox to protect the upstream IMAP service, and fall back to cached data if the source is briefly unreachable.
- Outbound that returns fast
- Your send call returns immediately with a server-generated Message-Id; a background worker delivers via Amazon SES with retries, so a transient hiccup doesn't lose your mail.
Your mailbox as agent tools
We run a native Model Context Protocol (MCP) server over a streamable-HTTP transport. Point an MCP-capable client — Claude Desktop, a remote agent — at the advertised endpoint and authenticate with a mailbox API key, and the mailbox's actions become tools your agent can call. No custom integration layer to build.
An API shaped for machines
Each API key is scoped to a single mailbox and carries one role — read, send, or admin — so you grant an agent only what it needs. Keys can expire (30 days, 90 days, 1 year, or never) and be revoked at any time. Mutating calls accept an Idempotency-Key so a retry doesn't double-execute, and every error comes back in a consistent envelope (type, code, message, param) modelled on Stripe's, easy to branch on. Over the rate limit, you get a 429 with Retry-After.
{
"error": {
"type": "invalid_request_error",
"code": "mailbox_not_found",
"message": "No mailbox matches that address.",
"param": "mailbox"
}
}A real error envelope — branch on code, not on a string.
Domain, DKIM and deliverability — handled
Order a domain and we place it on OVH, attach our default anycast nameservers, and configure it to host email. We publish the DKIM records automatically and poll Amazon SES until the sending identity is verified. MX and DKIM records are managed and reconciled, so drift gets fixed for you. Hard bounces and complaints are suppressed automatically, per mailbox, and bounce-rate guard rails protect the shared sending reputation. Your first-year domain registration is covered by your plan's domain credit.
Secure and EU-resident, by design
Each customer gets a dedicated encryption key. Mailbox SMTP/IMAP passwords, cached message bodies, and outbound bodies are all encrypted at rest with it, and underlying passwords are never returned by an API or written to logs. Add optional two-factor authentication (TOTP) with 10 recovery codes. Export a signed archive of your account at any time, and delete on a 7-day grace period that destroys your encryption key. Your mail lives on OVH; outbound is relayed via Amazon SES in eu-west-1 (Dublin).
Every plan ships the full platform
Pick the size, not the feature set. The Agent plan and the Swarm plan run the same engine — only the number of mailboxes, the monthly send volume, and the rate limit change.
- Native MCP server
- REST API
- Real-time inbound (?live=true)
- Scoped, expiring API keys
- Managed domain, DKIM & deliverability
- Per-customer encryption
- EU data residency
- Optional two-factor authentication
Give your agent the whole stack
EU data residencyPer-customer encryptionNative MCP