GoodSender vs Mailgun vs SendGrid vs Resend: an honest comparison from people who built one of them

Most “honest comparison” articles aren’t. They’re affiliate-driven, scored by writers who’ve never operated an MTA, or quietly weighted toward whoever’s paying for the listing.

This isn’t that. The team writing this co-founded SendGrid in 2009, built Laneful in 2025 for the largest email senders in the world, and is now building GoodSender. SendGrid is in the comparison below. So is Resend, the modern challenger. So is Mailgun, the long-running incumbent. We’ve sat on the receiving end of mailbox-provider escalations at 2 a.m. across all of these stacks and watched what actually moves deliverability and what’s marketing copy.

Yes, biased. The bias just happens to come from people who’ve operated the things they’re comparing, including the original one.

The short version (priced at 100,000 emails a month)

All pricing verified on 2026-04-20, including GoodSender’s. Re-check before relying on these numbers; rates change.

The rest of this piece is the part the comparison table can’t carry.

Where Mailgun is the right answer

Mailgun has the broadest feature surface of the four. Inbound routing, mature webhook infrastructure, validation APIs, a long-tenured deliverability team, EU/US region split. If you’re running a marketing program with complex segmentation, several inbound parsing endpoints, and an existing toolchain wired to their webhooks, ripping it out to save a few hundred dollars a month is rarely the right trade.

Where it costs you. $75 a month for 100K covers the full weight of the abuse-defense org, and there’s no consent layer, so list quality is your problem, not theirs. Reputation is shared across IP pools by tier. Clean senders pay for dirty ones, structurally. There’s no MCP server, hosted or local, so the AI-agent path is yours to build.

Where SendGrid is the right answer

We’re going to be even-handed about this. We co-founded SendGrid in 2009, watched it scale, watched it get acquired by Twilio, and watched what it became after we left. It is still the largest, most enterprise-integrated email API in the market. Brand-name comfort matters when the buyer is a procurement team. Existing apps already use it. The integrations and templates ecosystem is the deepest of any vendor in the comparison.

If your buyer cares about vendor maturity, your team has deep operational SendGrid knowledge from a previous role, or you’re sitting inside a Twilio-bundled relationship, the inertia is real and sometimes worth more than the savings.

Where it costs you. Per-email pricing with no consent layer, same shared-pool problem as the others. The free plan is gone (since May 2025), so the on-ramp hurts more than it used to. Setup is the most complex of the four. There’s no MCP story.

Where Resend is the right answer

Resend has the best developer experience of the four. The API is clean, the dashboard is well-designed, the documentation is genuinely a pleasure to read. If you’re a React or Next.js developer shipping a side project and you want something that feels modern out of the box, Resend is a defensible choice for the right reader.

Where it costs you. The price curve gets steep at scale. $35 a month for 100K is competitive. Roughly $650 a month at one million emails is the line item that ends the romance, and the tiers above that get worse. There’s no consent layer. The MCP server is local-only, which means it runs on your machine and not on the network, so anything that wants to send mail from a hosted agent (Claude desktop, n8n cloud, a Cursor session on someone else’s laptop) needs a workaround.

Where GoodSender is the right answer

If your primary need is my app needs to send email and I’d like to stop thinking about deliverability, this is the case we’re built for.

The Permission Loop means recipients confirm before you can mail them, which structurally eliminates the abuse surface that drives per-email pricing on the other three. The Engagement Check prunes silent recipients automatically and suppresses unsubscribes workspace-wide, so list quality is enforced by the platform rather than your discipline. A local MCP server is available so a Claude or local agent can send mail without you wiring a custom integration. The IP pool is pre-warmed and dedicated to GoodSender traffic, so new customers send at full volume from day one. Past 100K a month, you get automatic dedicated-IP allocation managed by GoodSender, no tier picking, no manual warm-up. Domain verification, including DKIM, SPF, and DMARC, takes about five minutes.

And the underlying pipe is Laneful, the same infrastructure already running the world’s largest senders, the company built by the SendGrid team after they left Twilio.

Where it costs you. It’s a permission-first model. If your business depends on cold outreach, scraped-list marketing, or warming up purchased data, GoodSender will make those workflows fail by design. That’s intentional. It’s also why we can be free for the volumes most senders ever reach. GoodSender is also explicitly not built for enterprise marketing teams running mass-broadcast campaigns. That audience belongs on Laneful, not here.

How to pick

If you’re sending mostly transactional from a real product to real users, try GoodSender. If you outgrow the model or hit a feature gap, look at Resend next for the DX win.

If you’re running a complex marketing program with deep webhook dependencies and you’re already on Mailgun, stay until a migration is genuinely worth it. Don’t move for the sake of moving.

If you’re inside an enterprise that uses Twilio products or has years of SendGrid templates and integrations, the migration cost is the bulk of your bill. Stay where you are unless the per-email price genuinely starts to bite, and re-evaluate when it does.

If you’re building an AI agent that sends email, Mailgun and SendGrid have no MCP story at all. Resend and GoodSender both ship local-only MCP servers, so the question is less about MCP availability and more about what model fits — Resend’s wider API surface, or GoodSender’s permission-gated, Markdown-native send.

If your decision is dominated by per-email cost at scale, model out the curve. The gap widens fast above 100K sends a month. At one million emails a month: Mailgun is roughly $700, Resend roughly $650, SendGrid roughly $700–750 (extrapolated — they publish 700K at $499 and skip to 1.5M, so the exact 1M figure depends on how their sales team tiers you), GoodSender roughly $9. The short version is that you’re paying for somebody else’s reputation, and the cost compounds.

How to actually decide

Three questions, in order:

Does my sending depend on cold outreach or scraped lists? If yes, none of GoodSender’s economics apply to you. Pick from Mailgun, SendGrid, or Resend on the basis of feature surface and price.

Am I locked in by integrations, procurement, or templates I’d have to rebuild? If yes, the migration cost dominates the per-email price. Stay where you are and revisit when something concrete breaks.

If neither of those is true, the question is just how much per-email pricing will cost you eighteen months from now. Model the curve at your projected volume, not your current one. That’s where the four products diverge most.