All posts

Building Calder: why I started a developer-first email API

Every project I shipped needed transactional email, and every time I wired up a provider I rebuilt the same three things. Calder is my attempt to build them once, properly.

EDaniel Oluwadare10 Nov 20262 min read
On this page · 4 sections

Calder started as an annoyance. Every project I shipped needed transactional email, password resets, receipts, notifications, and every time I wired up a provider I ended up rebuilding the same three things: templates, delivery tracking, and a dashboard that tells me what actually happened to the email I sent.

Email as an afterthought

Email is the only feature users judge by its absence. A broken checkout page gets reported. A password reset that never arrives just becomes a user who never comes back. Yet in most codebases email is the last thing wired up, usually with a provider SDK copy-pasted from a tutorial and a prayer.

I wanted the opposite: email as a first-class primitive, designed the way I'd design any other API. One call in, a predictable event stream out.

The first primitive

Calder's first primitive is a single POST: a template, a recipient, and variables. Everything else, rendering, sending, retries, webhooks, hangs off that one shape.

The system behind it is deliberately boring, too: an API app, a dashboard app, an async worker for sending and retries, and shared packages for the database, queue, providers and observability. Example curl first, so nothing depends on a published SDK:

bash
curl -X POST https://api.calder.dev/v1/emails \
  -H "Authorization: Bearer $CALDER_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "template": "password-reset",
    "to": "[email protected]",
    "variables": { "name": "Ada", "resetUrl": "https://…" }
  }'

One call. Rendering, retries and delivery events hang off it, and the events come back as webhooks, sent, delivered, bounced, complained.

Tip

Start with the boring primitive. If the core call feels inevitable, the rest of the product designs itself.

What I learned

  • Idempotency keys are not optional. Retries happen; double-sending a password reset is a bug users can see.
  • Webhooks need signatures and replay protection from day one, retrofitting trust is painful.
  • The dashboard is the product. Developers don't buy email sending; they buy knowing what happened.
  • TypeScript end to end pays for itself the first time a template variable changes shape.

Where it honestly stands

Calder is not "done", and I'd rather tell you that than pretend:

  • The API, dashboard, durable queue, drain loop, worker and SES integration are implemented and were manually verified end to end.
  • Billing isn't configured yet.
  • The SDK source exists, but I'm not claiming a published package, the HTTP quickstart is the path that works today.
  • SMTP is specified, not deployed.

Note

When your README has a "launch availability" section, keep it honest. The fastest way to lose a developer's trust is a feature list that's really a roadmap.

The bet is settled though: email deserves the same engineering respect as payments. Build it once, build it properly.

Did this help?

Send it to someone fighting the same bug.

DO

Daniel Oluwadare · Emerald

Software engineer on TestFlow, and the person writing everything here. I build in public and post what I learn.

Say hello

Keep reading

Building Calder: why I started a developer-first email API · Emerald