---
title: "The best email API for AI agents in 2026, eight options compared"
description: "Eight ways to give an AI agent an email address, compared on the question that matters most, what happens when a hostile message arrives. We run one of them. Every number comes from the vendor's own pages, linked."
canonical: "https://machinemail.dev/blog/best-email-api-for-ai-agents-2026"
date: "2026-11-04"
---

# The best email API for AI agents in 2026, eight options compared

> Eight ways to give an AI agent an email address, compared on the question that matters most, what happens when a hostile message arrives. We run one of them. Every number comes from the vendor's own pages, linked.

- From: MachineMail Team <blog@machinemail.dev>
- Date: 2026-11-04
- Tags: comparison, agents
- Source: https://machinemail.dev/blog/best-email-api-for-ai-agents-2026
- Reply: blog@machinemail.dev

We run one of the products on this list, so here is the deal. We put MachineMail first and we tell you why. Every number about another product comes from that product's own pricing or docs page, read on 4 November 2026, and each one links to its source. Where a rival is the better choice, we say so. Judge the reasoning, not the ranking.

The questions we asked are the ones an agent actually poses. Can it get an inbox of its own, by code, without anyone editing DNS? When mail arrives, does the agent get the message or a notification that a message exists? Who checks the message before the agent reads it, and what happens to phishing and to instructions aimed at the agent? Is the conversation threaded? Can the agent wait for a one-time code? Where does the data live? And what does the free plan really include?

Five products were built inbox-first for agents: MachineMail, AgentMail, OpenMail, Mailtrap Agent Inbox and Nylas Agent Accounts. Resend Inboxes is an inbox product on a sending platform. Three more are transactional senders that can receive: Postmark, Mailgun and Amazon SES. That split matters more than anything else below. Prices are in US dollars unless marked as euros.

## At a glance

| Product     | Built for agents | Own inbox by API | Address with no DNS | Mail checked before the agent reads it | Agent-safe text | Threads | Wait for a code | EU option    | Free custom domain |
| ----------- | ---------------- | ---------------- | ------------------- | -------------------------------------- | --------------- | ------- | --------------- | ------------ | ------------------ |
| MachineMail | ✓                | ✓                | ✓                   | ✓ trust report, every plan             | ✓               | ✓       | ✓ 60 s          | ✗            | ✓ one              |
| AgentMail   | ✓                | ✓                | ✓                   | partial: after delivery, Labs, US      | ✗               | ✓       | ✗               | Enterprise   | ✗                  |
| OpenMail    | ✓                | ✓                | ✓                   | partial: sender rules; MCP skips spam  | ✗               | ✓       | ✗               | ✓ every plan | ✗                  |
| Mailtrap    | partial          | ✓                | ✓                   | ✗                                      | ✗               | ✓       | ✗               | ✗            | ✓ one              |
| Nylas       | ✓                | ✓                | ✓ trial domain      | partial: spam and DNSBL                | ✗               | ✓       | not documented  | ✓ every plan | ✗                  |
| Resend      | partial          | ✓                | ✗                   | ✗                                      | ✗               | ✓       | ✗               | ✗            | ✓ three            |
| Postmark    | ✗                | ✗                | ✓ hash address      | partial: spam score                    | partial         | ✗       | ✗               | ✗            | ✓                  |
| Mailgun     | ✗                | ✗                | partial: sandbox    | partial: spam score, off by default    | partial         | ✗       | ✗               | ✓ region     | ✓ one              |
| Amazon SES  | ✗                | ✗                | ✗                   | partial: verdicts, advisory            | ✗               | ✗       | ✗               | ✓ five EU    | n/a                |

"Partial" means the product does something in that column but not what the column asks. Each product's section says exactly what, with a source.

## 1. MachineMail

Every agent gets its own address with one call, free, on machinemail.dev or your domain. The part we built the company around is what happens next. Every inbound message gets a [trust report](https://machinemail.dev/docs/trust/trust-report) before it is stored and before any event fires: SPF, DKIM and DMARC alignment, prior correspondence, your allow and block lists, link and attachment checks, hidden-text detection and a classifier for spam, phishing and instructions aimed at the agent. The report gives one of five levels, and mail that comes out `quarantined` never reaches the agent: it gets the `QUARANTINE` label instead of `INBOX`, fires `message.received.quarantined`, stays out of lists, search, MCP reads and `wait_for_message`, and a person reviews it in the dashboard.

What the agent reads is [`trust.content.text`](https://machinemail.dev/docs/trust/content): the sender's new words, with quoted history, signatures, disclaimers, text hidden from human readers and invisible Unicode removed. The raw body is kept. The hosted [MCP server](https://machinemail.dev/docs/agents/mcp) has 16 tools built for mail an agent did not ask for, including `wait_for_message`, which blocks for up to 60 seconds until a sign-up code or magic link arrives. Events come over signed webhooks, a WebSocket, Server-Sent Events that resume from `Last-Event-ID`, and MCP Events for received, delivered and bounced mail. Inboxes can carry a `ttl_seconds` and delete themselves, a verified domain can be catch-all, and an agent with no key can [register itself](https://machinemail.dev/docs/agents/auth-md) into a sandbox inbox that a person later claims with a six-digit code. The API serves AgentMail's `/v0` shape, so AgentMail's TypeScript SDK works with a different base URL.

Where we are behind: no SOC 2 or other certification, data stored in the US only, no SDK of our own, no IMAP or SMTP, no calendar, no open tracking, no text extraction from PDFs or Office files, a smaller MCP tool set and integrations list than AgentMail, OpenMail or Resend, and no dedicated IPs.

Price: Free at $0 with 3,000 messages a month, 1 GB and one custom domain. Pro $9, Team $49, Scale $199 a month. Inboxes are free on every plan. See [pricing](https://machinemail.dev/pricing).

Pick it when your agent reads mail from strangers and acts on it: support, sign-ups and codes, agents that reply on their own.

## 2. AgentMail

The other product built around the inbox rather than the pipe, and the one with the broadest surface: dedicated inboxes, threading, a hosted MCP server with 47 tools plus calendar, IMAP and SMTP access, open tracking, TypeScript and Python SDKs, SOC 2 Type II, and integrations from LangChain to the OpenAI and Claude agent SDKs. We compare it in detail in [MachineMail vs AgentMail](https://machinemail.dev/compare/agentmail).

On inbound safety AgentMail has Agent Armor, which scans for prompt injection after delivery. In the default Monitor mode the message arrives and `message.received` fires as usual, then a verdict follows, usually within seconds; a Hold mode keeps `review` messages from agents until released. It is in Labs, free while there, US only, and an organization admin turns it on ([Agent Armor](https://docs.agentmail.to/agent-armor.md)). No cleaned, agent-safe text field is documented, and the MCP server has no wait tool ([MCP](https://docs.agentmail.to/integrations/mcp.md)). Agents can sign up through `/v0/agent/sign-up`, then a person verifies by email ([agent onboarding](https://docs.agentmail.to/agent-onboarding.md)). EU cloud is an Enterprise plan feature ([pricing](https://www.agentmail.to/pricing)).

Price: Free at $0 with 3 inboxes and 3,000 emails a month and no custom domain. Developer $20 with 10 inboxes and 10 custom domains, Startup $200 with 150, Enterprise on request. Extra inboxes $2 a month ([pricing](https://www.agentmail.to/pricing)).

Pick it when you need SOC 2 Type II, IMAP, a calendar, an official Python SDK, or the integrations list, and your agent's inbound mail is low-risk or screened by you.

## 3. OpenMail

Built inbox-first in Vilnius, with EU data residency on every plan including Free; only Nylas matches that, as a US-or-EU choice, and OpenMail is the one product here that is EU-hosted by default ([pricing](https://openmail.sh/pricing)). Each inbox has its own keys and sending reputation, pods separate tenants as isolated sub-accounts, and keys scope down to a single inbox so an injected instruction in one agent's mail cannot reach another's ([authentication](https://docs.openmail.sh/api-reference/authentication.md)). Attachments arrive with `parsedText` pulled from PDFs, Office files, CSVs and images through OCR ([attachments](https://docs.openmail.sh/concepts/attachments.md)), and a WebSocket replays up to 100 missed events with a webhook fallback ([WebSockets](https://docs.openmail.sh/concepts/websockets.md)). Full comparison: [MachineMail vs OpenMail](https://machinemail.dev/compare/openmail).

Inbound safety is address-based. Sender rules are evaluated before delivery, and denied mail is dropped with no bounce and no quarantine view; rules are "address-based only" with "no filtering on subject, body content, or attachment type" ([sender rules](https://docs.openmail.sh/concepts/sender-rules.md)). Over the REST API, mail from an allowed address is delivered as received, with no trust level or cleaned text; the hosted MCP server's `list_unread_threads` does skip spam, marketing, bounces and machine-generated mail ([MCP](https://docs.openmail.sh/integrations/mcp.md)), but nothing in the payload says why. There is no wait tool and no agent self-signup yet ([llms.txt](https://openmail.sh/llms.txt)).

Price, in euros: Free at €0 with 3 inboxes and 3,000 emails a month, no custom domain. Pro €9 a month plus usage, with 10 inboxes, 10,000 emails and a spend cap. Enterprise on request ([pricing](https://openmail.sh/pricing)).

Pick it when your data must stay in the EU on every plan, you want attachment text in the event, or you run many tenants who must not hurt each other.

## 4. Mailtrap Agent Inbox

Mailtrap is the email testing tool most developers already have, and since June 2026 it receives mail too: "Create an inbox, get a @inbound-mailtrap.io address, and any email sent to it lands in the API plus fires a webhook" ([changelog](https://feedback.mailtrap.io/changelog)). Inboxes are created per agent through the API, on a hosted address or your own domain, every message is parsed to JSON with attachment download URLs, replies are threaded, and an MCP server lets Claude or Cursor create inboxes and reply in thread ([Agent Inbox](https://mailtrap.io/agent-inbox/)). Inbound is included in every Email API plan, Free included ([pricing](https://mailtrap.io/pricing.md)). Agent Inbox is Mailtrap's general Inbound Email product plus an MCP server, not a separate product.

The webhook carries event IDs and metadata, and the agent fetches the message ([webhooks](https://docs.mailtrap.io/inbound-email/webhooks.md)). On safety, Mailtrap's own inbound roundup says there is "no spam verdict in the payload (under development for Inbound at the time of writing)" ([Mailtrap blog](https://mailtrap.io/blog/best-inbound-email-api/)); its inbound page does count "spam caught" in analytics, so some filtering runs, but nothing about it reaches the agent. The message schema has no authentication results, and allow lists, block lists, quarantine and cleaned text are not in the docs we checked. The MCP server runs locally only, replies from hosted inboxes are capped at 20 in total, and a person must create the account ([AI onboarding](https://docs.mailtrap.io/for-ai-agents/ai-onboarding.md)). Mailtrap holds SOC 2 Type II and ISO 27001; data is hosted in the US; "EU data residency is in development" ([security](https://mailtrap.io/security-and-compliance/)).

Price: Free at $0 with 4,000 emails a month shared between sending and receiving, 150 a day, three days of retention and one domain. Basic $15 for 10,000 ([pricing](https://mailtrap.io/pricing.md)).

Pick it when you already test on Mailtrap and want agent inboxes and an MCP server from the same account.

## 5. Nylas Agent Accounts

Nylas has long connected apps to people's Gmail and Outlook. Agent Accounts, launched in June 2026 and now generally available, is the reverse: one `POST /v3/connect/custom` provisions a real mailbox with a calendar, and it works with every existing Nylas endpoint and SDK ([overview](https://developer.nylas.com/docs/v3/agent-accounts.md)). A free `*.nylas.email` trial domain "is usable the moment you create it, with no DNS setup" ([quickstart](https://developer.nylas.com/docs/v3/getting-started/agent-accounts.md)); for your own domain, Nylas generates the records and you publish them, with DMARC left to you ([DNS setup](https://developer.nylas.com/docs/v3/agent-accounts/dns-provider-setup.md)). Webhooks carry the message body, and the hosted MCP server has US and EU endpoints ([MCP](https://developer.nylas.com/docs/dev-guide/mcp.md)). Compliance is the strongest here: SOC 2 Type II, ISO 27001, ISO 27701, HIPAA with a BAA on paid plans, and a US or EU region on every plan ([security](https://www.nylas.com/security/), [pricing](https://www.nylas.com/pricing/)).

Inbound safety is a mail provider's: spam detection with DNSBL checks and a sensitivity dial (basic on Free), SPF, DKIM and DMARC checks, and on paid plans typed allow and block lists whose block rules reject "at the SMTP level" ([policies](https://developer.nylas.com/docs/v3/agent-accounts/policies-rules-lists.md)). There is no trust level or quarantine, no cleaned text in the payload (a separate clean-messages endpoint strips script and style tags), and prompt injection is left to you: "Treat all email and calendar content as untrusted input" ([agent security](https://developer.nylas.com/docs/v3/getting-started/agent-security.md)). No TTL and no catch-all; a wait tool is not in the docs we checked, which point agents at webhooks or polling `GET /messages` ([supported endpoints](https://developer.nylas.com/docs/v3/agent-accounts/supported-endpoints.md)).

Price: Free at $0 with 3 Agent Accounts, 3,000 emails a month and 20 sends an account a day, on the trial domain only. Essentials $15 with 10 accounts and one custom domain. Pro $49 with 50 accounts and 10 domains ([pricing](https://www.nylas.com/pricing/)).

Pick it when the agent needs a calendar as much as an inbox, you need HIPAA or ISO 27001, or you are already on Nylas.

## 6. Resend Inboxes

The developer favourite for sending, and deservedly: Forge sends from IP blocks Resend owns, there are SDKs in eight languages, 85+ MCP tools by Resend's count, and SOC 2 Type II. Inboxes launched on 22 October 2026 on top of that platform, with threads, labels, search, drafts and a built-in triage agent that drafts replies for a person to send ([Inboxes](https://resend.com/products/inboxes)). Full comparison: [MachineMail vs Resend](https://machinemail.dev/compare/resend).

An inbox address must be on a domain you own, with DKIM and SPF verified and an inbound MX record, or forwarding from an existing mailbox; the request returns `422` otherwise ([create inbox](https://resend.com/docs/api-reference/inboxes/create-inbox.md)). The `inbox.email.received` webhook carries metadata only and the body is fetched from the API ([event types](https://resend.com/docs/webhooks/event-types.md)). The thread email response has no authentication field; SPF, DKIM and DMARC verdicts are documented only on the separate Receiving API ([thread email](https://resend.com/docs/api-reference/inboxes/get-thread-email.md), [received email](https://resend.com/docs/api-reference/emails/retrieve-received-email.md)). Spam and phishing classification, allow and block lists, quarantine, cleaned text, a WebSocket and a wait tool are not in the docs we checked. "A human must create the Resend account" ([AI onboarding](https://resend.com/docs/ai-onboarding.md)). Stored data is held in the US ([security](https://resend.com/security)), and email data is retained for 30 days on Free, Pro and Scale ([quotas](https://resend.com/docs/knowledge-base/account-quotas-and-limits.md)).

Price: Free at $0 with 3,000 emails a month, 100 a day, three domains and 3 inboxes; received mail counts against the same quota. Pro $20 for 50,000, with extra inboxes at $1 a month ([pricing](https://resend.com/pricing)).

Pick it when you already send through Resend, you want people and an agent working the same inbox, or you want an SDK in Go, Rust, Java or .NET.

## 7. Postmark

Postmark's sending reputation is what others measure against, and its inbound parsing is thorough: the webhook posts the full message as JSON with text and HTML bodies, headers, base64 attachments, a `StrippedTextReply` and SpamAssassin scoring ([inbound webhook](https://postmarkapp.com/developer/webhooks/inbound-webhook)). Inbound is on the Free, Pro and Platform plans, not Basic ([pricing](https://postmarkapp.com/pricing)). Addresses come free: a hash address at `inbound.postmarkapp.com`, plus-addressing into `MailboxHash`, or a wildcard inbound domain ([inbound server](https://postmarkapp.com/developer/user-guide/inbound/configure-an-inbound-server), [inbound domains](https://postmarkapp.com/developer/user-guide/inbound/inbound-domain-forwarding)).

There is no inbox object or thread model; messages are stored for 45 days by default (7 to 365 on Pro and up) and searchable by sender, recipient, subject and tag ([messages API](https://postmarkapp.com/developer/api/messages-api)). Safety is a spam score: set a threshold and Postmark blocks mail above it, and block individual senders or domains ([inbound blocking](https://postmarkapp.com/developer/user-guide/inbound/configure-inbound-blocking)). Phishing detection, allow lists and a wait tool are not in the docs we checked; the only cleaned field is `StrippedTextReply`, which strips quoted history from English replies only. Inbound webhooks are not signed; Postmark recommends basic auth and IP allow-listing ([webhooks overview](https://postmarkapp.com/developer/webhooks/webhooks-overview)). Postmark has no SOC 2 of its own and no EU hosting ([EU privacy](https://postmarkapp.com/eu-privacy)), and new accounts are reviewed by hand before they can send outside their own domains ([account approval](https://postmarkapp.com/support/article/1084-how-does-the-account-approval-process-work)). The MCP server is local and has no inbound tools ([MCP](https://postmarkapp.com/blog/the-postmark-mcp-server-one-year-later-from-4-tools-to-24)).

Price: Free at $0 with 100 emails a month, sent and received together, and 10 domains. Basic $15 for 10,000 without inbound; Pro $16.50 with inbound; Platform $18 ([pricing](https://postmarkapp.com/pricing)).

Pick it when delivery is everything and receiving is an occasional, parsed bonus.

## 8. Mailgun

Of the transactional senders, Mailgun has the most flexible inbound: routes pair a filter with an action, filters take regular expressions with captures, and the action forwards to a URL, stores the message briefly (the docs say up to three days; the pricing table lists one day on Foundation and up to seven on Scale), or stops ([routes](https://documentation.mailgun.com/docs/mailgun/user-manual/receive-forward-store/routes), [filters](https://documentation.mailgun.com/docs/mailgun/user-manual/receive-forward-store/route-filters)). The posted payload includes `body-plain`, `stripped-text` and `stripped-signature`, and since September 2026 can arrive as JSON ([receiving via HTTP](https://documentation.mailgun.com/docs/mailgun/user-manual/receive-forward-store/receive-http), [releases](https://www.mailgun.com/releases/)). An EU region keeps message data in the EU ([API overview](https://documentation.mailgun.com/docs/mailgun/api-reference/api-overview)), and Mailgun holds SOC 2 Type II and ISO 27001 ([security](https://www.mailgun.com/security/)).

No inbox, no threading, and search is the events log. Spam filtering is SpamAssassin, off by default, surfaced as `X-Mailgun-Sflag` and `X-Mailgun-Sscore` headers ([spam filter](https://documentation.mailgun.com/docs/mailgun/user-manual/receive-forward-store/spam-filter)); inbound allow and block lists and phishing detection are not in the docs we checked; `stripped-text` removes quoted parts and the signature block, nothing more. The MCP server is local only ([MCP](https://documentation.mailgun.com/docs/mailgun/mcp)). Routes are plan-gated: one on Free, five on Basic, full access from Foundation ([pricing](https://www.mailgun.com/pricing/)). The sandbox domain every account gets can receive on one route, for testing only ([sandbox domains](https://documentation.mailgun.com/docs/mailgun/user-manual/domains/domains-sandbox)).

Price: Free at $0 with 100 emails a day, one domain and one route. Basic $15 for 10,000 with five routes. Foundation $35 for 50,000 with full inbound routing ([pricing](https://www.mailgun.com/pricing/)).

Pick it when you already send with Mailgun and want to route inbound mail by rule, or you need EU inbound from a sender.

## 9. Amazon SES

The cheapest way to move email at volume. Receiving needs a verified domain, an MX record and a receipt rule set whose actions land the raw MIME in S3 or an SNS notification, or invoke a Lambda with the metadata ([receiving concepts](https://docs.aws.amazon.com/ses/latest/dg/receiving-email-concepts.html)). SES scans for spam and viruses, checks SPF, DKIM and DMARC, and blocks senders on the Spamhaus and SES block lists, but "doesn't take any actions on received email based on the results" ([receiving concepts](https://docs.aws.amazon.com/ses/latest/dg/receiving-email-concepts.html)); the verdicts arrive as attributes for your code ([notification contents](https://docs.aws.amazon.com/ses/latest/dg/receiving-email-notifications-contents.html)). Receiving works in 22 regions, six of them in Europe, five in the EU plus London ([endpoints](https://docs.aws.amazon.com/general/latest/gr/ses.html)), and SES is in scope for SOC, PCI and HIPAA programs ([compliance](https://docs.aws.amazon.com/ses/latest/dg/compliance-validation.html)).

No inbox, no threading, no search, no IMAP ([receiving concepts](https://docs.aws.amazon.com/ses/latest/dg/receiving-email-concepts.html)); SNS notifications carry mail up to 150 KB and S3 up to 40 MB ([quotas](https://docs.aws.amazon.com/ses/latest/dg/quotas.html)). New accounts sit in a sandbox that caps sending at 200 messages a day until production access is granted; AWS gives an initial response within 96 hours and may ask for more ([production access](https://docs.aws.amazon.com/ses/latest/dg/request-production-access.html)). AWS's own agent skill "does not include Mail Manager or email receiving" ([agent setup](https://docs.aws.amazon.com/ses/latest/dg/agent-setup-guide.html)). Phishing detection and cleaned text are not in the docs we checked.

Price: à la carte, inbound is $0.10 per 1,000 emails plus $0.09 per 1,000 chunks of 256 KB, and outbound $0.10 per 1,000; new accounts start on the Essentials plan at $0.16 per 1,000 sent, and new AWS customers get up to $200 in credits ([pricing](https://aws.amazon.com/ses/pricing/)).

Pick it when you have an AWS team, a lot of mail, and your own code for everything above the SMTP layer.

## Which one

If the agent only sends: Resend or Postmark, and stop reading. If it must receive and act on what it reads, you want an inbox-first product, and then the choice is about what you trust the inbox to check. Mail from strangers that the agent acts on, codes and sign-ups, support: MachineMail. Data that must stay in the EU on a free plan, or many tenants to isolate: OpenMail. SOC 2, IMAP, calendar or the biggest integrations list: AgentMail. A calendar, HIPAA or ISO 27001: Nylas. Already testing on Mailtrap: Mailtrap. Already sending with Resend and want a shared team inbox: Resend. Rule-based routing from a sender with an EU region: Mailgun. An AWS team with volume: SES.

## Frequently asked questions

### What is the best email API for an AI agent that sends and receives?

One built around the inbox: MachineMail, AgentMail, OpenMail, Mailtrap Agent Inbox or Nylas Agent Accounts, or Resend Inboxes on your own domain. Transactional APIs receive at the domain level and leave the inbox, the threading and the state to you.

### Which email API for AI agents checks inbound mail before the agent reads it?

MachineMail, on every plan, with a trust level and a cleaned text field on every message. AgentMail's Agent Armor scans after delivery, in Labs, US only. Postmark and Mailgun can block above a spam threshold you set. Nylas filters spam into a junk folder and, on paid plans, rejects listed senders at the SMTP level, but exposes no verdict field. SES hands you verdicts to act on yourself. OpenMail filters by sender address. Mailtrap and Resend Inboxes document no inbound classification today.

### Which email API for AI agents is hosted in the EU?

OpenMail on every plan. Nylas lets you pick a US or EU region on every plan. Mailgun and Amazon SES offer EU regions. AgentMail's EU cloud is Enterprise only. MachineMail, Mailtrap, Resend and Postmark store data in the US.

### Which includes a custom domain on the free plan?

MachineMail includes one. Resend includes three verified domains. Postmark, Mailgun and Mailtrap include domains on Free as well. AgentMail, OpenMail and Nylas do not; Nylas gives you a trial subdomain instead.

### What does an agent inbox cost?

MachineMail: inboxes are free on every plan, and the Free plan covers 3,000 messages a month. AgentMail: 3 free, then $20 a month for 10 and $2 an inbox after that. OpenMail: 3 free, then €9 a month for 10 and €1 an inbox after that. Resend: 3 on every plan, $1 an inbox beyond that on paid plans. Nylas: 3 free, then $15 a month for 10. Mailtrap: inbound is bundled with the sending plan.

### How do I stop prompt injection by email?

Do not put a raw email body into the model's context, decide from a trust signal the sender cannot forge, and keep anything that fails that check away from the agent. MachineMail does the three for you: `trust.content.text`, `trust.level` and quarantine. On the other products you build it from spam scores, authentication results and your own filters; AgentMail's Agent Armor and Nylas's block rules each cover part of it.

## Sources

All checked on 4 November 2026.

AgentMail: [pricing](https://www.agentmail.to/pricing), [Agent Armor](https://docs.agentmail.to/agent-armor.md), [MCP](https://docs.agentmail.to/integrations/mcp.md), [agent onboarding](https://docs.agentmail.to/agent-onboarding.md).

OpenMail: [pricing](https://openmail.sh/pricing), [sender rules](https://docs.openmail.sh/concepts/sender-rules.md), [authentication](https://docs.openmail.sh/api-reference/authentication.md), [attachments](https://docs.openmail.sh/concepts/attachments.md), [WebSockets](https://docs.openmail.sh/concepts/websockets.md), [MCP](https://docs.openmail.sh/integrations/mcp.md), [llms.txt](https://openmail.sh/llms.txt).

Mailtrap: [Agent Inbox](https://mailtrap.io/agent-inbox/), [pricing](https://mailtrap.io/pricing.md), [changelog](https://feedback.mailtrap.io/changelog), [webhooks](https://docs.mailtrap.io/inbound-email/webhooks.md), [best inbound email API](https://mailtrap.io/blog/best-inbound-email-api/), [AI onboarding](https://docs.mailtrap.io/for-ai-agents/ai-onboarding.md), [security](https://mailtrap.io/security-and-compliance/).

Nylas: [Agent Accounts overview](https://developer.nylas.com/docs/v3/agent-accounts.md), [quickstart](https://developer.nylas.com/docs/v3/getting-started/agent-accounts.md), [DNS setup](https://developer.nylas.com/docs/v3/agent-accounts/dns-provider-setup.md), [policies, rules and lists](https://developer.nylas.com/docs/v3/agent-accounts/policies-rules-lists.md), [supported endpoints](https://developer.nylas.com/docs/v3/agent-accounts/supported-endpoints.md), [agent security](https://developer.nylas.com/docs/v3/getting-started/agent-security.md), [MCP](https://developer.nylas.com/docs/dev-guide/mcp.md), [security](https://www.nylas.com/security/), [pricing](https://www.nylas.com/pricing/).

Resend: [Inboxes](https://resend.com/products/inboxes), [create inbox](https://resend.com/docs/api-reference/inboxes/create-inbox.md), [event types](https://resend.com/docs/webhooks/event-types.md), [thread email](https://resend.com/docs/api-reference/inboxes/get-thread-email.md), [received email](https://resend.com/docs/api-reference/emails/retrieve-received-email.md), [AI onboarding](https://resend.com/docs/ai-onboarding.md), [security](https://resend.com/security), [quotas](https://resend.com/docs/knowledge-base/account-quotas-and-limits.md), [pricing](https://resend.com/pricing).

Postmark: [pricing](https://postmarkapp.com/pricing), [inbound webhook](https://postmarkapp.com/developer/webhooks/inbound-webhook), [inbound server](https://postmarkapp.com/developer/user-guide/inbound/configure-an-inbound-server), [inbound domains](https://postmarkapp.com/developer/user-guide/inbound/inbound-domain-forwarding), [messages API](https://postmarkapp.com/developer/api/messages-api), [inbound blocking](https://postmarkapp.com/developer/user-guide/inbound/configure-inbound-blocking), [webhooks overview](https://postmarkapp.com/developer/webhooks/webhooks-overview), [EU privacy](https://postmarkapp.com/eu-privacy), [account approval](https://postmarkapp.com/support/article/1084-how-does-the-account-approval-process-work), [MCP](https://postmarkapp.com/blog/the-postmark-mcp-server-one-year-later-from-4-tools-to-24).

Mailgun: [routes](https://documentation.mailgun.com/docs/mailgun/user-manual/receive-forward-store/routes), [route filters](https://documentation.mailgun.com/docs/mailgun/user-manual/receive-forward-store/route-filters), [receiving via HTTP](https://documentation.mailgun.com/docs/mailgun/user-manual/receive-forward-store/receive-http), [spam filter](https://documentation.mailgun.com/docs/mailgun/user-manual/receive-forward-store/spam-filter), [sandbox domains](https://documentation.mailgun.com/docs/mailgun/user-manual/domains/domains-sandbox), [API overview](https://documentation.mailgun.com/docs/mailgun/api-reference/api-overview), [MCP](https://documentation.mailgun.com/docs/mailgun/mcp), [releases](https://www.mailgun.com/releases/), [security](https://www.mailgun.com/security/), [pricing](https://www.mailgun.com/pricing/).

Amazon SES: [receiving concepts](https://docs.aws.amazon.com/ses/latest/dg/receiving-email-concepts.html), [notification contents](https://docs.aws.amazon.com/ses/latest/dg/receiving-email-notifications-contents.html), [quotas](https://docs.aws.amazon.com/ses/latest/dg/quotas.html), [production access](https://docs.aws.amazon.com/ses/latest/dg/request-production-access.html), [endpoints](https://docs.aws.amazon.com/general/latest/gr/ses.html), [compliance](https://docs.aws.amazon.com/ses/latest/dg/compliance-validation.html), [agent setup](https://docs.aws.amazon.com/ses/latest/dg/agent-setup-guide.html), [pricing](https://aws.amazon.com/ses/pricing/).

MachineMail facts come from the [MachineMail docs](https://machinemail.dev/docs) and were checked against the code the same day. Found something out of date? Reply to this post and we will fix it.
