Agent-run cold email platform vs. hand-assembling the vendors yourself
This deepens the summary on the Compare page into a concrete walk-through of what a coding agent actually has to do, and hold in context, to hand-assemble cold-email infrastructure from raw vendor APIs — versus driving a facade that already holds that state. Factual only: no competitor named or disparaged, no fabricated deliverability numbers.
SPEC.md §12 in the repository, sourced 2026-07-09). Time estimates are structural (how many sequential steps and how many weeks a process mechanically requires), not a benchmark of any specific agent or competitor. No deliverability, open-rate, or reply-rate comparison is made anywhere on this page.
Assembling it yourself: what actually has to happen
A coding agent asked to "run cold email outreach end-to-end" that decides to hand-roll it needs to sequence through all of the following, correctly, before a single compliant email can send:
- Registrar account and API key (e.g. Porkbun or Namecheap) — a separate signup, a separate credential to hold, a separate rate limit and error surface to handle.
- Lookalike-domain generation logic — brand-derived candidates (
tryacme.com,getacme.com) that read as legitimate rather than as phishing, built and maintained by you or your agent. - Domain purchase — roughly $11-15/yr per .com/.net domain at public wholesale pricing.
- Manual DNS per domain — SPF, DKIM, DMARC, and reverse DNS/PTR records, each configured correctly, because a single misconfigured record now gets mail rejected at the SMTP level under 2026 Gmail/Yahoo rules (see the deliverability guide).
- A mailbox/warmup vendor account (e.g. Inboxkit) — another signup, another credential, roughly $5.5-6/mailbox/month all-in at public pricing, plus a multi-week warmup ramp your agent has to track state for across sessions.
- A sequencing/reply engine — scheduling, per-mailbox send-cap enforcement, bounce/reply/unsubscribe detection, threading — either self-built or a separate OSS/vendor tool with its own setup.
- Compliance logic, implemented correctly, every time — CAN-SPAM opt-out handling, a working suppression list, RFC 8058 one-click unsubscribe, physical-address footer injection — none of which is enforced by any of the vendors above; it's on you or your agent to build and not skip.
- Ongoing per-mailbox monitoring — bounce rate, spam-complaint rate, and warmup health, checked and acted on regularly, for as long as the campaign runs.
None of this is unusually hard individually. What makes it costly is that a coding agent doesn't persist state between sessions — the vendor credentials, the current warmup day per mailbox, the suppression list, the domain DNS status, all have to be re-derived or re-fetched from three-plus separate systems every time the agent is invoked to do something with this pipeline, for as long as the campaign is active (which, given multi-week warmup, is measured in weeks-to-months, not one session).
The same flow through a stateful facade
| Layer | DIY, self-integrated | agent-cold-email |
|---|---|---|
| Credentials to hold | Registrar API key + mailbox-vendor API key + (optionally) sequencer credentials — three-plus separate secrets, three-plus auth models. | One bearer token, one auth model, resolved fresh per call. |
| Domains + DNS | Generate candidates, buy each, configure SPF/DKIM/DMARC/rDNS per domain, verify manually. | One setup_infrastructure call; DNS is verified automatically before a mailbox is provisioned on it. |
| Mailboxes + warmup | Separate vendor account, manual per-mailbox warmup tracking across weeks of sessions. | Provisioned and warmed as part of the same call; infrastructure_status reports current state on demand — no state to re-derive. |
| Sequencing + replies | Self-built or separate tool: scheduler, caps, bounce/reply/unsubscribe detection, threading. | Built into the engine; suppression and unsubscribe enforced automatically. |
| Compliance guardrails | Implemented by you or your agent, correctly, every campaign, with no server-side backstop if a step is skipped. | Enforced server-side by default (suppression, send caps, complaint auto-pause, footer address injection) — not dependent on the agent remembering. |
| Re-wiring cost per session | Every new agent session touching this project re-derives vendor state from three-plus systems. | One token; usage, quota, warmup state, and billing live durably in the platform. |
Where DIY is honestly the better call
- You need a specific vendor or dedicated-IP setup this platform doesn't yet support — it currently targets Inboxkit-class self-serve vendors (see
SPEC.md§13 in the repository). - One-off, very low volume — a handful of contacts, once, makes three vendor signups a real option, not obviously worse.
- You already have warmed mailboxes and a sequencer from prior work — there's nothing here to add.
Wholesale economics behind the numbers above
Domains ≈ $11-15/yr (.com/.net via Porkbun, public pricing); mailboxes + warmup ≈ $5.5-6/mailbox/mo (Inboxkit-class vendor, all-in, public pricing); Stripe processing ≈ 2.9% + 30¢. Fully-loaded cost per mailbox, including domain amortization and the normal 8-18%/month domain-burn replacement cycle (see why that's normal, not a failure), comes to roughly $7/mo — full sourcing in docs/research/vendor-tos-economics-2026-07-09.md in the repository. This platform's retail pricing targets a 2.5-3x margin over that wholesale cost.
No deliverability comparison against any named competitor exists on this page or anywhere on this site.
Related
- How to run cold email with your AI coding agent — the setup path this comparison is measured against.
- Compare (summary) — the shorter version of this page.
- Cold email deliverability in 2026.