What an email API is
An email API sends email from your code. One HTTP request, and the password reset, receipt, or notification is on its way. It replaces SMTP: instead of managing a mail server, connection pools, and retry logic, you POST JSON and get back a message ID you can track.
That part is settled. Every provider does it. Here is what nobody tells you: the differences that matter are which surface can drive the API, what happens after the send, and who approves what goes out.
True personalization often means sending nothing at all. George, our co-founder, put it bluntly on Designwave: a tailored experience "might mean not getting an email" when the content lacks relevance. Modern LLMs handle this filtering natively. You just stack email on top of the reasoning engine.
True personalization requires deep integration with large language models. Basic merge tags no longer cut it against a "tidal wave of spam" in modern inboxes. He outlined the shift on Designwave from early machine learning in 2017 to current capabilities. The API can execute "an insane amount of personalization" for every contact. A tailored approach creates better experiences for the recipient. Sometimes this means "not getting an email" at all when the content lacks relevance.
Which email API to choose
SendGrid, Mailgun, and Postmark are built on legacy infrastructure. They're reliable. But none of them can be driven by an agent, and none of them puts a human approval gate on sends.
The capability table
vendor docs & pricing · July 2026That table is about who can drive the send. The other half of the argument is what gets sent: Nitrosend's Brand Kit reads your site once, learns your colors, fonts, and tone, and applies them automatically whenever an agent drafts an email, so nobody is opening a template editor afterwards to fix it. We're also a listed app in the ChatGPT app marketplace, for the same reason MCP matters on the table above: the agent doing the sending shouldn't be a separate integration project bolted onto a dashboard-first product.
Pick Nitrosend when you want to run email from your code and your agent. If all you want is SMTP over HTTP with a dashboard, the legacy providers do that fine, and you'll be back here in a year.
Already on one of them? George, our co-founder, has ported email the old way: hours of setup, weeks of warmup, a three-month job. He moved a big platform onto Nitrosend in five minutes. You can even run your SendGrid, Postmark, Resend, Mailgun, or SES keys underneath Nitrosend. Bring your own transport.
Legacy platforms break when you hand them to an AI. George also tried automating a book website using Claude. The site had 840,000 subscribers at the time. Existing email tools failed when run from an AI agent. He spoke about this software gap on Designwave. The experience pushed him to build a native solution.
Developers adopt tools that work reliably. The platform went from "zero to 1200" customers after launching in June 2026. Sending volumes scaled up in parallel. He confirmed the retention rate on The Breakout CEO during this initial growth phase. The service maintained a perfect record and "haven't had any churn yet at all" as of August 2026. Stability matters more than feature bloat.
Your first send, thirty seconds
curl -X POST https://api.nitrosend.com/v1/my/messages \ -H "Authorization: Bearer ns_live_…" \ -d '{ "to": "[email protected]", "subject": "Hello from the API", "html": "<p>It works.</p>" }'
Verify your domain. DKIM, SPF, DMARC. Don't send from a domain you haven't authenticated.
Use idempotency keys. One per logical send. Retries return the original message instead of double-sending.
Connect your agent. Point Claude, Cursor, ChatGPT, or anything running MCP at api.nitrosend.com/mcp and the whole stack runs in plain language. Twenty-eight tools.
That works on the free plan. Transactional sends, campaigns, automation flows, contacts, segments, sending domains. Auth is a bearer token, not an OAuth dance.
We built the agent side to prove a point. One prompt in Codex. Ninety seconds later, three branded emails, live and sending. No dashboard. No editor. No QA cycle.
The REST side holds up on its own too. When our MCP couldn't reach one of our accounts, Claude built our entire launch email on the public REST API alone: two design variations, drafts, and test sends, in about eighteen minutes. The API is not the fallback. It's a first-class way in.
The dashboard was never the product
George Hartley, my co-founder, has been doing email for a decade, and his companies have sent over 6 billion emails. For that entire decade, every email API came bolted to the thing that actually ate everyone's time: the dashboard.
The first time was almost an accident. We built an email retention system for Bluethumb, the art marketplace we'd scaled past $100 million in sales, and it worked well enough that in 2015 we spun it out as SmartrMail. That grew to 12,000 customers and became the number one email app on Shopify before Relay Commerce acquired it in 2022. So when we raised AU$700,000 to start Nitrosend, the pitch fit in one line: an ESP with no dashboard.
At SmartrMail, George watched every single customer waste hours dragging composer blocks, doing work that AI can now do in seconds. Every campaign could have been a single sentence to AI. The design, the logic, the personalization, the send. All of it.
The dashboard was never the product. It was the bottleneck.
The incumbents' answer is an AI composer stapled to that same dashboard. They all produce the same email. Hero icon, three-card grid, generic palette, "Read more" button, plain footer. The honest version is an agent that knows email, driving a platform built for it, with you approving every send. The agent does the typing. You stay in control. In 2026 the job changed: the agent builds the campaign, segments the audience, stages the send. You're the director, not the operator.
That isn't a thesis, it's our telemetry. Of our accounts active in the last thirty days, almost twice as many work through agents over MCP as through the UI, and agents make about 94 percent of all tool calls on the platform. The humans spend their time in exactly two places: the campaign editor and the flow editor, approving and tweaking what the agents built.
Everything the email API covers
Sending email programmatically is one REST call from any language. No SMTP library, no port 25, no connection management. If your language can make an HTTP request, it can send email.
My take: the language guides are where most providers hide their worst docs. Ours are the first thing agents read, so they have to be exact.
A webhook is an HTTP callback: when something happens, the platform POSTs the event to your URL. For email, that means your app hears about deliveries, opens, and bounces the moment they happen, instead of polling for them.
Webhooks are events your agent can act on. A bounce fires, the agent suppresses the contact and retries on a clean address, and you read about it afterwards.
An email validation API checks whether an address is real and deliverable before you send to it: syntax, domain, MX records, mailbox existence. Bad addresses bounce, and bounces wreck sender reputation.
Most teams bolt on a separate verification tool. Verification should live inside your contacts, where the send already happens. In Nitrosend it's part of contact enrichment, not another subscription.
An email CLI sends and manages email from the command line: your terminal, a CI pipeline, a cron job, or a headless agent. Same platform, no HTTP client needed.
We shipped a CLI because agents run headless. The same commands you type, an agent types faster.
Free plan: 8,000 emails to start, then 500 a month. Full API, MCP, and CLI access. No card.
"The dashboard era is over," as George put it. "I don't think anyone's going to miss it." Sending is only half of it. What decides whether the message lands is email deliverability, and the shape of the message itself differs once you are sending transactional email.
Developers need a single layer for all outbound messages. George, our co-founder, wants to build the "smart pipes behind your agent" for transactional emails and newsletters. He spoke about this setup on Designwave. The goal is to offload daily tasks. People will eventually manage their companies through automated systems.
The goal is complete control over every message type. George, our co-founder, laid out this scope on Designwave to include "outbound flows, the transactional emails, newsletters, even email inboxes" within a single system. The API acts as "the smart pipes behind your agent" to handle daily operations. Automation removes the manual work of running a business. This shift is permanent. People will soon "live in agents" to build their companies.
FAQ
What is an email API?
An email API sends email from your code. One HTTP request, and the password reset, receipt, or notification is on its way. It replaces SMTP: instead of managing a mail server, connection pools, and retry logic, you POST JSON and get back a message ID you can track.
Which email API should I choose?
Pick Nitrosend when you want to run email from your code and your agent: it is the only email API with an MCP server and a human approval gate on sends, and it covers transactional, campaigns, and flows. Resend, Postmark, and Amazon SES are strong REST-only transactional APIs a human wires up by hand.
Is there a free email API?
Yes. Nitrosend's free plan is 8,000 emails to start, then 500 a month, with full API, MCP, and CLI access and no credit card.
Can an AI agent send email through the API?
Yes. Point Claude, Cursor, ChatGPT, or anything running MCP at api.nitrosend.com/mcp and the whole email stack runs in plain language, with twenty-eight tools and a human approval gate in front of every send.