Nitrosend vs Mailforge.

TL;DR

Mailforge is cold email infrastructure. It sells domains at $14 a year and mailbox slots at $3 a month billed annually, ten slots minimum, on its own distributed pool, with DMARC, SPF, DKIM, and a tracking domain configured for every domain you add. The mailboxes arrive pre-warmed and export into whatever sequencer you run. If your job this week is standing up a hundred sending mailboxes, it does that job cheaply.

These are two of the three layers a sending stack runs on, not two versions of one product. Mailforge is the layer cold mail leaves from, and Nitrosend is the platform the rest of your email runs on: campaigns, flows, segments, and transactional, with the workflow surface driven as MCP tools rather than screens. Most companies that buy one find they needed both, on different domains. If what you want is more mailboxes, buy Mailforge.

1. Leave the cold stack alone. Domains, mailboxes, and the sequencer stay where they are, nothing in this migration touches them, and no domain has to be transferred.
2. Authenticate your real domain in Nitrosend. DKIM and SPF go on the domain your customers already recognise, kept deliberately apart from the domains cold mail is sent from.
3. Import the permissioned list, and only that list. nitro_import_contacts takes email, first and last name, and custom properties. Cold prospects don't belong in this import.
4. Move one message type first. Send a receipt or a password reset with nitro_send_message and watch it for a week before a campaign follows it, then rebuild anything scheduled as a flow with nitro_compose_flow.
5. Decide what carries the permissioned reputation. BYO sending keys on Pro and above route mail through Amazon SES, Resend, Postmark, Mailgun, or SendGrid, on a domain and IP you already warmed. Mailforge's pool isn't one of them and isn't meant to be: it's a shared pool carrying cold traffic, which is the last reputation you'd want your receipts riding on.

• Unlimited contacts on every plan, including Free, so a list that grows without being mailed never moves the bill.
• Campaigns, flows, and segments in the same product as transactional, with the two streams kept deliverability-isolated.
• Permissioned email has somewhere to live, which is the half of the program an infrastructure provider doesn't cover.
• MCP-first is an architecture, not an integration: every capability is an API endpoint and an MCP tool before it's a screen.
• The agent surface reaches the work itself, so campaigns, flows, contacts, templates, and segments are tools rather than a read-only slice.
• Deliverability, billing, tests, and approvals are tools too, and they're the four that usually stop an automation dead.
• Transactional sends stay on the REST API, because a send that has to be fast shouldn't wait on a tool call.
• A REST API and a CLI sit alongside the MCP, so nobody is forced through an agent to get work done.
• A free tier you can run a real program on, in a category where evaluating infrastructure usually starts at a purchase.
• BYO sending keys on Pro and above route mail through Amazon SES, Resend, Postmark, Mailgun, or SendGrid, on the domain and IP you already warmed.
• BYO keys make the platform decision reversible, where switching usually means changing sending reputation on the same day.

A cold email stack is three purchases, not one.

Cold outbound runs on three layers, and they're sold separately. The first is the domains and the mailboxes, which is what Mailforge sells. The second is the sequencer that writes the steps and pushes the sends out through those mailboxes. The third is the platform the rest of your email runs on: the receipts, the password resets, the lifecycle mail, and the campaigns that go to people who asked for them.

The three get shopped as though they substituted for each other. That's how a buyer ends up weighing a sequencer's seat price against a mailbox provider's $3 a mailbox, buying two of the three, and never noticing the gap. The pattern we see is a company with a hundred sending mailboxes and a healthy outbound number, whose order confirmations still leave from a script somebody wrote in an afternoon and nobody has opened since. The first two layers already run themselves by the time anyone buys them: the mailbox pool provisions on request, and the sequencer sends on a schedule nobody watches. The permissioned layer is the one still worked by hand, screen by screen, which is exactly what makes it the layer worth putting behind an agent command.

Naming the layers still doesn't tell you how many mailboxes you need. Below a dozen or so, a ten-slot minimum is the wrong shape for the problem, and the honest answer for a lot of companies is one authenticated domain with a platform on top of it and no cold layer at all.

Keep the cold domains away from the rest of your email.

A cold program sends from domains bought for the purpose, mailed to people who never asked, and a share of the recipients complain. The domains a cold program sends from are being spent. That's what they're for. Bounces and complaints on those domains are the running cost of doing outbound, and it's a cost worth paying when the outbound works.

The domain your company actually trades under can't absorb any of that. Receipts, password resets, and the campaigns your customers opted into have to land, and they land because that domain carries a clean authentication history and a complaint rate that barely moves. Running both from one domain is the mistake that takes six months to show up and another six to undo, which is why the separation is worth setting up before it's urgent rather than after.

Separating the domains doesn't separate the brand, though. A prospect who complains about your cold mail still associates it with your company, so the split protects deliverability rather than reputation in the wider sense.

If the permissioned half of your email is the part still being done by hand, that's the part we built for. Nitrosend runs campaigns, flows, segments, and transactional from the domain you actually trade under, with the workflow surface reachable as MCP tools, and contacts are unlimited on every plan. Start on the free tier, point one message type at it, and read the pricing ladder once you know your real volume.

Does Nitrosend give me cold email mailboxes like Mailforge does?

No. Nitrosend is sender-side: it sends from a domain you authenticate with DKIM and SPF, and it doesn't provision, host, or read mailboxes. If the job is standing up sending mailboxes across a batch of domains, buy Mailforge for it.

Can I run Mailforge and Nitrosend at the same time?

Yes, and it's the arrangement we'd recommend. Cold outbound leaves the Mailforge domains and mailboxes, everything permissioned leaves the domain your customers already know, and the two reputations stay apart on purpose. Neither side has to know about the other, because they're separate subscriptions doing separate jobs.

Does Mailforge send the email, or do I still need something else?

Mailforge is infrastructure. It provisions domains and mailboxes, sets up the DNS, and hands them to whatever sending software you run, so the composing, the sequencing, and the sending are always a second purchase. That's a deliberate design choice on their side rather than a gap: it's what lets the mailboxes work with any sequencer.

What does Nitrosend do that Mailforge has no equivalent for?

Campaigns, automation flows, segmentation over an owned contact list, templates, and transactional sending, with the workflow surface reachable as MCP tools rather than as a dashboard with an API bolted under it. Mailforge sits a layer below all of that, which is why most companies running both find the two bills answer different questions.