Platform

How a stream meets its objective.

Every stream runs through the same loop: declare what it is for, give it its own lane, hold its policy per provider, and prove the result message by message.

Say what each stream is for, and what success means.

A stream is a kind of email with one job: sign-in codes, receipts, product updates, a campaign. Every message carries its stream, and every stream carries an objective, a lane, a pool, and its policies.

  • Mail type on every message

    Set transactional or promotional per message, through the API or over SMTP. It decides the lane before the message is queued.

    Reference
  • Tags and metadata

    Label messages by stream, product, or tenant (otp, receipt, spring-launch) so every event, webhook, and report can be cut the same way.

  • Stream objectives

    Each stream’s objective is agreed with your deliverability engineer during onboarding and tracked per mailbox provider in reporting.

Stream plan

Sign-in & OTP

Agreed with your engineer
Sent as
mail_type: transactional · tag: otp
Objective
Delivered in under 5 s, p95, at every provider
Lane
Priority. Never queued behind bulk mail
Pool
auth-eu-1 · dedicated, kept warm
Policy
Fast retry on deferral · suppression enforced at accept
Reported
p95 from send to the provider's 250 OK, per provider, weekly with your engineer
Other streams:TransactionalLifecyclePromotional

POST /v1/transmissions

Idempotency-Key: otp-7731-ada

202 accepted
{
  "recipients": [{ "email": "ada@example.com" }],
  "content": { "template_id": "tpl_signin_code" },
  "options": {
    "mail_type": "transactional",
    "tags": ["otp"]
  }
}

Give every stream a lane of its own.

A campaign spike should never slow a sign-in code, and a neighbour’s complaints should never decide your inbox placement. Streams, pools, and tenants are kept apart.

  • Transactional and promotional lanes

    Separate queues with separate priorities. A campaign never sits in front of a password reset, and never shares its reputation.

  • IP pools

    Create pools, assign IPs, route streams and tenants to them. Reputation boundaries you draw, not a shared pool you inherit.

    Reference
  • Sending domains

    Create, verify, and manage domains through the API, with SPF and DKIM readiness for every brand, product, or tenant.

    Reference
  • Tenant isolation

    Per-tenant credentials, rate limits, webhook policy, and database pools. Built for platforms sending on behalf of their own customers.

  • Vetted network

    Every sender is reviewed before it sends. A smaller, cleaner network is part of keeping your reputation yours.

POST /v1/ip-pools

scope manage:ip-pools

201 created
{
  "name": "auth-eu-1",
  "ips": ["203.0.113.10", "203.0.113.11"],
  "routing": {
    "streams": ["transactional"],
    "tags": ["otp"]
  },
  "warmup_policy": "auth-eu-1-ramp"
}
Pools are routing targets for streams and tenants. Warmup policies attach to them.

Hold each stream’s policy, provider by provider.

Write the rule once. The pipeline enforces it at accept time and at send time, including at 2am when a campaign spikes and nobody is watching.

  • Pacing policies

    Budgets per provider, per stream, per pool. Deferrals slow the lane before the provider does it for you. Activate, pause, and inspect budgets through the API.

    Reference
  • Warmup policies

    Ramp schedules for new domains and pools, with progress tracking and slowdowns on deliverability signals. Repeatable across tenants.

    Guide
  • Suppressions

    Bounces, complaints, and unsubscribes enforced when the message is accepted. JSON upsert or bulk import for migrations and hygiene.

    Reference
  • Provider-aware retries

    A delay queue for deferrals. Retry timing follows the provider’s response instead of hammering a server that asked you to wait.

    Guide
  • Platform guardrails

    Provider circuit breakers, per-tenant limits, queue-depth caps, and fair batching for promotional traffic.

Warmup · promo-eu-2 → gmail.comPromotional
1k10k100k300kplan · 300k/day by day 30Gmail 421sheld 2 days, then resumedtoday · day 14cap 13k/dayday 1day 15day 30
Today’s cap
13k / day
Gmail deferrals
0.4%
Schedule
+2 days, by policy

POST /v1/pacing-policies

scope manage:pacing

200 ok
{
  "name": "promo-gmail-ramp",
  "type": "warmup",
  "scope": { "stream": "promotional", "pool": "promo-eu-2" },
  "provider": "gmail.com",
  "schedule": [
    { "day": 1,  "max_per_hour": 200   },
    { "day": 7,  "max_per_hour": 2000  },
    { "day": 14, "max_per_hour": 12000 },
    { "day": 21, "max_per_hour": null  }
  ],
  "on_deferral":  { "slow_by_pct": 10, "cooldown_s": 900 },
  "on_complaint": { "pause_if_rate_gt": 0.001 }
}

Measure it against its objective, message by message.

Every attempt and every provider response becomes a typed event. That is how a stream’s result is proven, and how a miss is explained down to the provider and the reply.

  • Stream scorecards

    Each stream against its own objective, by mailbox provider, week by week. Built on the deliverability reporting below.

  • Message and transmission events

    Per-recipient timelines with every attempt and response, for support cases, audits, and reconciliation.

    Reference
  • Signed webhooks

    Delivery, deferral, bounce, complaint, open, click, unsubscribe. HMAC-signed, with secret rotation and a test endpoint.

    Reference
  • Bounce explorer

    Failures grouped by provider, category, and response signature. A block shows up as a pattern across thousands of messages, not as a ticket.

  • Engagement tracking

    Opens and clicks with signed tokens. Security scanners and link prefetchers are filtered out of your numbers.

  • Deliverability and engagement reporting

    Summaries, time series, and an explorer for ad-hoc questions, on an analytics store built for wide queries.

GET /v1/events/message/{message_id}

delivered in 2.1 s
[
  { "type": "accepted", "ts": "14:22:08.012",
    "tags": ["otp"] },
  { "type": "queued",   "ts": "14:22:08.015",
    "lane": "transactional", "pool": "auth-eu-1" },
  { "type": "attempt",  "ts": "14:22:08.089", "n": 1,
    "mx": "outlook-com.olc.protection.outlook.com" },
  { "type": "deferral", "ts": "14:22:08.913",
    "smtp_code": 451, "enhanced_code": "4.7.650" },
  { "type": "attempt",  "ts": "14:22:09.640", "n": 2,
    "mx": "outlook-com.olc.protection.outlook.com" },
  { "type": "delivery", "ts": "14:22:10.101",
    "smtp_code": 250, "reason": "2.6.0 Queued mail for delivery" }
]
A sign-in code, deferred once and retried within a second, still inside its five-second objective.

Send over the API or SMTP, into the same pipeline.

HTTP for the services you own, SMTP for the ones you do not. Both land in the same lanes, policies, and evidence.

  • Transmissions API

    One endpoint that accepts a message and returns a transmission ID. Acceptance and delivery are separate events.

    Reference
  • SMTP submission

    Authenticated SMTP for legacy applications and third-party tools, with credentials scoped per app or environment.

  • Idempotency keys

    Send the same request twice with the same key and the second is a no-op. Retry freely on timeouts and deploys.

    Reference
  • Templates

    Store content once and reference it at send time. Smaller payloads, consistent branding.

    Reference
  • Scoped API keys

    send:* keys for services that send, manage:* keys for the people and tooling that configure.

    Reference

How the pipeline is built.

An asynchronous pipeline where acceptance, delivery, and evidence are separate stages with separate guarantees.

  1. Edge

    Authentication, scope checks, per-tenant rate limiting, and idempotency enforcement before a message is accepted.

  2. Intake and lanes

    Accepted messages are queued at once into separate transactional, promotional, and delay lanes.

  3. Workers and MTA

    Delivery workers enforce pacing and warmup per provider and pool, then hand off to the MTA layer that holds the SMTP conversation.

  4. Events and analytics

    Every attempt and response becomes a typed event, dispatched to webhooks with retries and written to the analytics store.

See it on one of your streams.

A parallel run on one stream tells you more than any feature list. Tell us your domain, your current provider, and which stream matters most, and an engineer will plan it with you.

What happens next

  1. We review the domain

    Every sender is vetted before it sends. A small, clean network is part of the product.

  2. We provision the environment

    Scoped keys, SMTP credentials, DNS records for SPF and DKIM, a webhook endpoint to test against.

  3. You move one stream

    Usually sign-in codes or receipts, measured against its objective from the first message.

The first call is with an engineer, not a sales script. We ask about your streams before we talk about us.