What mails.ai is, what we ship today, and where we’re headed.
Mails.ai builds email for AI agents. Not marketing email. Not newsletters. Email that agents send, receive, and parse — programmatically. Sign-up is self-serve and live today: free tier, no card.
Mails.ai is email infrastructure built for AI agents — a programmatic email API where the inbox primitive is the design center, not an afterthought of a send-first transactional product. Every agent gets its own address, with reply parsing, identity and reputation baked in.
Three phases, in order:
Phase 1 (Q3 2026) — Launch. Free 3K emails + 3K inbound / Pro $20 (50K emails + 50K inbound) / Scale $99 (250K + 500K), billed monthly or yearly through Stripe. Structured reply events, six-category prompt-injection scanner, per-agent reputation graph, MCP server, TypeScript and Python SDKs.
Phase 2 (Q4 2026) — Receiver-side verification API + browser extension, multi-vendor failover hardening (documented runbook for Postmark / SparkPost / Brevo with a ~4-hour DNS swap path).
Phase 3 (Q1 2027) — Per-agent budgets + auto-throttle. Recipient-side reputation verification. Agent-to-agent direct-delivery protocol.
Plans and prices in full on the pricing page.
Three things we leave out on purpose, each dated the day we wrote it down.
The mails.ai team<support@mails.ai>
We do not ship marketing campaigns, segmented broadcasts, or visual template editors. Agent infrastructure is the focus. SendGrid and others handle that surface well; we do not compete with them.
The mails.ai team<support@mails.ai>
All tiers are self-serve. No SSO/SAML/procurement-only tier. No “contact us for Enterprise.” The pricing page is the price.
The mails.ai team<support@mails.ai>
Same self-serve friction floor as Resend and the rest of the transactional category. Abuse is handled architecturally: a per-agent reputation classifier scores every send, suppression runs at send time, and a per-sender complaint cron auto-suspends at 0.3%.
Search every message an agent sent, read whole conversations, and watch events arrive as they happen, all from one dashboard.
The SDKs, the MCP server and the docs are public. Questions about any of them go to one address, read by our team.
The API is live and self-serve today, with its architecture documented from the start.
