SparkPost

Promptfree, for everyone, and the only version there is

Copy the prompt for Claude CodeCopy the prompt for CodexCopy the prompt for Cursor

Copy

Do not build the sending infrastructure. Build the analytics and control layer above it, which is genuinely valuable and entirely yours.

**The reason to rent delivery is that inbox placement is not a code problem.** Where a message lands depends on your domain's history, your complaint rate and the receiving provider's opinion of you — none of which you can write.

STACK
- Node 20+ with Fastify, server-rendered HTML
- SQLite through better-sqlite3, WAL mode
- A relay provider for delivery only
- Caddy in front

THE LAYER WORTH OWNING
- templates, versioned, with the version recorded on every message sent
- messages and events as usual: queued, sent, delivered, bounced, complained
- suppressions, permanent, checked before every send
- **domains: id, name, spf_ok, dkim_ok, dmarc_policy, checked_at — verified by your own DNS check on a schedule.** A DKIM record removed during an unrelated DNS change is a silent deliverability collapse, and nobody notices for a week

THE ANALYTICS THAT ACTUALLY MATTER
- Bounce rate split into hard and soft, and by receiving domain. **The split by receiving domain is the diagnostic**: if Gmail is fine and one provider is bouncing everything, that is a reputation problem with one provider and it is fixable
- Complaint rate by campaign and by day, with an alarm at one in a thousand
- Delivery latency, which catches a relay problem early
- **Open tracking is misleading and worth reconsidering.** Privacy proxies pre-fetch images, so opens are inflated by an unknown amount that varies by audience. Report clicks, which are real, and treat opens as a weak signal at best

BOUNCE HANDLING DONE PROPERLY
- Parse the bounce classification: a hard bounce suppresses permanently, a soft bounce retries with backoff and suppresses after several consecutive failures
- A mailbox-full bounce is not an invalid address; treating it as one loses real subscribers
- Complaints from feedback loops suppress immediately and permanently

LIST HYGIENE, WHICH IS THE REAL DELIVERABILITY LEVER
- **Stop sending to people who have not opened or clicked in six months.** This is the single most effective thing anybody does for deliverability and it costs nothing
- Confirmed opt-in on every list, with the timestamp and source recorded
- Never import a purchased list. It ends your sending reputation and possibly your provider account

STAYING PORTABLE
- Talk to the relay through one small adapter. Switching provider should be one file and a set of credentials — **that is the whole point of owning this layer**

WHAT MATTERS MOST
Bounce classification, list hygiene and a scheduled DNS check on your own authentication records. Those three do more for inbox placement than any vendor feature.