Trotta Mail vs MachineMail: two inboxes that check mail before an agent reads it
Both check inbound mail before it is stored and keep dangerous mail from agents. Pick MachineMail for free inboxes, low metered prices and an AgentMail-compatible API. Pick Trotta Mail if you want outgoing mail screened too, held mail withheld from every API key by default, and humans' mailboxes in the same product.
Updated
MachineMail and Trotta Mail are close. Both check every inbound message before it is stored, both give the model a cleaned view of the text, and both keep dangerous mail from agents until a human looks. The differences are in what is checked, which routes withhold held mail, and outgoing mail.
Checked against MachineMail's public docs, security page and pricing page on November 20, 2026. Prices are as listed on their pricing page on 2026-11-20.
Side by side
| Trotta Mail | MachineMail | |
|---|---|---|
| Built for | AI agents and humans. Humans get a mailbox in the browser. | AI agents, with a dashboard where a team reviews and takes over threads. |
| When incoming mail is screened | While the sending server is still connected, before the message is accepted. Known-malicious mail is refused at the door. | Their docs state every message is checked between parsing and storage. Refusal during delivery is not stated. |
| What is checked | Prompt injection in the body and in documents, phishing, spoofing, lookalike domains, deceptive links and disguised programs. | Authentication, sender history, hidden text, links, attachments, and scoring for spam, phishing and prompt injection. |
| Held mail and human review | Held outside the inbox and withheld on every API key route, whatever the key's scopes, unless a human gave it guard:release. | Quarantined and withheld from agents until a human reviews it. MCP tools withhold it. Over REST, a key limited to the agent-safe view gets only the sender. |
| Outgoing checks | Screened for secrets, sensitive data and replies steered to a new address. Optional human approval. | Their docs state outbound messages have no trust report. Content screening of outgoing mail: not found in their public docs. |
| Sign-in codes | Sealed from every API key in a human's mailbox. Parsed for the agent in its own inbox. | Parsed for the agent. Keeping codes from a key: not found in their public docs. |
| MCP and SDKs | MCP with an API key or OAuth sign-in. TypeScript and Python SDKs, REST, webhooks, live events. | Hosted MCP with OAuth or an API key. REST compatible with AgentMail's SDKs, a CLI in early access, webhooks, live events. |
| Own domains | Paid plans: up to 3 on Developer, 25 on Startup. | 1 on Free, 10 on Pro, 50 on Team, 250 on Scale. |
| IMAP and SMTP access | No. | Not found in their public docs. |
| Price to start | Free: 2 inboxes, 1,000 emails a month. | Free: inboxes free, 3,000 messages a month, 100 a day. |
| Plans as listed | Developer, $20 a month: 5 inboxes, 5,000 emails. Startup, $200 a month: 50 inboxes, 50,000 emails, 5 seats. Enterprise: custom. | Pro, $9 a month: 7,500 in, 2,500 out. Team, $49: 45,000 in, 15,000 out. Scale, $199: 200,000 in, 65,000 out. Unlimited seats on paid plans. |
Where MachineMail is the better choice
- Inboxes cost nothing. MachineMail does not charge per inbox: up to 1,000 on its shared domain and as many as you like on your own. Trotta Mail plans cap inboxes.
- Lower prices and more seats. Pro is $9 a month with unlimited seats. Spam and quarantined mail are never billed.
- An AgentMail-compatible API. Its
/v0API follows AgentMail's paths and fields, so an existing AgentMail client moves by changing the base URL and key. - Team features. Single sign-on and passkeys for the dashboard, thread owners so a human can take over from an agent, and export of any inbox as mbox or JSONL.
- Throwaway inboxes. Ephemeral inboxes that expire on their own, and a catch-all domain that turns any address into an inbox.
Where Trotta Mail is the better choice
Trotta Mail is the better choice when the protection has to hold whatever key an agent is given. A Trotta Mail key reads the safe view unless a human grants more, and held mail stays withheld even from a key that reads full mail.
- Checked during delivery. Known-malicious mail is refused while the sending server is still connected. It never reaches the inbox, and an agent's key cannot read what was refused.
- Outgoing mail is screened. A secret in a message an agent sends is refused, and replies steered to a new address are caught. Each connection can ask before every email or send drafts only.
- Documents are read. Text in PDF, Word, Excel, PowerPoint and OpenDocument files is screened, including text formatted to be invisible.
- Humans have their own mailbox. Sign-in codes and reset links in a human's mailbox are sealed from every key, and an agent there writes drafts for the human to send.
No filter catches everything, Trotta Mail included. A cleverly reworded scam with no technical tell can still be delivered, which is why uncertain mail is held for a human.
Switching from MachineMail
Switching from MachineMail maps one verdict model onto another.
- Create the inbox with one call:
POST /v1/agents/signupwith no account, orPOST /v1/inboxeswith a key. - Point your webhook, event stream or WebSocket at Trotta Mail, or call
GET /v1/inboxes/{id}/wait. - Map trust levels to Trotta Mail's verdict:
actionisdeliver,flag,holdorreject, andlevelissafe,cautionordanger. - Plan inbox counts against Trotta Mail's plan limits, since inboxes are not free there.
- On a paid plan, connect your domain: one click at providers that support Domain Connect, or five DNS records.
MachineMail is a trademark of its owner. This page is Trotta Mail's own comparison, based on MachineMail's public documentation as of November 20, 2026.
Questions
Is Trotta Mail a MachineMail alternative?
Yes. Both give agents inboxes that check mail before storing it and keep dangerous mail from them. Trotta Mail also screens outgoing mail and gives humans a sealed mailbox; MachineMail has free inboxes and lower prices.
Does MachineMail screen mail before an agent reads it?
Yes. Its docs state every inbound message is checked between parsing and storage, and that quarantined mail is withheld from agents until a human reviews it.
Does MachineMail screen outgoing mail?
Its docs state outbound messages have no trust report, and no content screening of outgoing mail was found in its public docs on November 20, 2026.
Which costs less?
MachineMail, as listed on November 20, 2026: inboxes are free and paid plans start at $9 a month. Trotta Mail's plans start at $20 a month for 5 inboxes.
Sources
- MachineMail docs: Trust layer
- MachineMail security
- MachineMail pricing
- MachineMail about
- MachineMail llms.txt (product summary)
- MachineMail docs, all pages (MCP tools, API keys, CLI, permissions)
- MachineMail changelog
- Trotta Mail API summary for agents
- Trotta Mail docs: security model, attachment text, plans and limits