# What still applies when an agent sends your email *Published: 2026-08-24* An agent can draft, segment, and send faster than a coordinator. It does not get a courtesy lane at Gmail. Mailbox providers score **domains, IPs, and complaint rates**. They do not score whether a human clicked Send. If the agent ships a promotional body through the transactional IP, you inherit the same problem you would have created by hand, only sooner. This is a companion to the [email deliverability guide](/blog/email-deliverability), not a replacement. That post covers reputation, authentication, and list hygiene. This one is about the mistakes that show up when software is the one hitting send. ## Key takeaways - Mailbox providers score domains, IPs, and complaint rates. They do not care whether a human clicked Send. - [CAN-SPAM](https://www.ftc.gov/business-guidance/resources/can-spam-act-compliance-guide-business) still uses the [primary-purpose test](https://www.ecfr.gov/current/title-16/chapter-I/subchapter-C/part-316/section-316.3). "The model wrote it" is not a defense. - Give the agent a promotional identity for campaigns. Keep [transactional](/blog/transactional-emails) mail on a separate transactional-purpose domain and, if you can, a separate IP. - Read the first unattended campaign against the source brief. Every figure should appear in a doc you own. ![Architecture diagram of an agent reaching Brew over MCP to compose, send, and read results.](/blog/brew-mcp-architecture.png) *The agent is another client on the same sending identity. The inbox still sees your domain.* ## The law does not mention agents [CAN-SPAM](https://www.ftc.gov/business-guidance/resources/can-spam-act-compliance-guide-business) applies to commercial messages. The [primary-purpose test](https://www.ecfr.gov/current/title-16/chapter-I/subchapter-C/part-316/section-316.3) is about the subject line and the body, not about who typed them. If the agent writes a "your invoice is ready" subject and a launch offer in the body, the message is commercial. You still need an honest subject, a physical address, and a working unsubscribe. "The model wrote it" is not a defense. The same is true in reverse. A password reset the agent triggered on a real security event is still transactional, as long as the body stays on that event. ## What Gmail and Yahoo actually require [Gmail's sender requirements](https://support.google.com/a/answer/81126) apply once you are in the bulk-sender bucket. The page's current threshold is **5,000** or more recipients in a day to personal Gmail. Marketing mail then needs: - SPF, DKIM, and alignment with DMARC on a `p=none` or stricter policy. - One-click unsubscribe per [RFC 8058](https://www.rfc-editor.org/rfc/rfc8058.html), `List-Unsubscribe` and `List-Unsubscribe-Post`. - Spam rates under **0.1%**, and never at **0.3%**. Transactional mail is exempt from the one-click rule. It is not exempt from authentication, and it is not exempt from spam complaints if you load the template with a sale. [Yahoo](https://senders.yahooinc.com/best-practices/) asks high-volume senders to put marketing and transactional on **different IPs**, and preferably different domains. An agent that "just uses the default sender" will collapse that split the first week you let it run unattended. ## What to put in the agent's reach Give the agent a **promotional** sending identity for campaigns and automations. The domain you warm, the IP you expect complaints on, the unsubscribe headers you have tested. Keep **transactional** mail on a separate identity. In Brew that identity is a domain with a transactional sending purpose, and the mail itself is a published automation with a trigger. The agent can compose the template. The send path per event should still be `POST /v1/automations/triggers/{triggerEventId}/fire` with a JSON payload, not "send this draft to the list." Do not let the agent pick the From address from a pile of verified domains "to see what lands." That is how you train two reputations at once and understand neither. ## Review before the first unattended send The failure mode is not a typo. It is a fluent paragraph that cites a number you never published, or a subject that looks like a receipt. Before you let an agent send without a queue: 1. Read the first campaign against the source brief. Every figure should appear in a doc you own. 2. Send it to a seed list that includes Gmail, Yahoo, and Outlook. 3. Confirm the one-click header is present on promotional mail, and left off transactional mail on purpose. 4. Confirm the physical address is in the promotional footer. Cory Nelson at Harmonya told [Noah Adelstein](https://thegtmengineer.substack.com/p/how-to-turn-email-marketing-into) he did that first-campaign read before he trusted the loop. That is the correct order. ## The short version Speed is not a new channel. It is the same channel with less time to notice you mixed a sale into a receipt. Authenticate the domain, split the IPs, tell the agent which identity is which, and keep a human on the first send. The brief that stops invented numbers is in [how to brief an email agent](/blog/how-to-brief-an-email-agent). The Harmonya first-campaign read is in [how Harmonya turned a newsletter into a lifecycle program](/blog/harmonya-lifecycle-email). To contact us about this article, write [Brew support](/support). ## How to use this page **Agent-sent email** is still email. Mailbox providers do not give models a courtesy lane. **The primary-purpose test** is how US law decides if the message is commercial. **One-click unsubscribe** is a header requirement for bulk commercial mail, including RFC 8058. For example, a receipt that spends most of the body selling a plan is still a campaign. | Move | Why it matters | | --- | --- | | Split marketing from transactional | Folder names in the ESP do not decide the law | | Authenticate the domain | The agent does not skip SPF, DKIM, or DMARC | US commercial mail still follows the [FTC CAN-SPAM guide](https://www.ftc.gov/business-guidance/resources/can-spam-act-compliance-guide-business) and [16 CFR 316.3](https://www.ecfr.gov/current/title-16/chapter-I/subchapter-C/part-316/section-316.3). Bulk sending still has to meet [Gmail's sender requirements](https://support.google.com/a/answer/81126). To contact us about this article, use the support link above. ## Frequently asked questions ### Do agents get different inbox rules? No. [Gmail](https://support.google.com/a/answer/81126) and [Yahoo](https://senders.yahooinc.com/best-practices/) score the sending identity. An agent is another client on your domain. ### Does transactional mail skip one-click unsubscribe? Gmail's bulk-sender page exempts transactional mail from the one-click header. It does not exempt it from authentication or from spam complaints if you load the template with a sale. One-click itself is [RFC 8058](https://www.rfc-editor.org/rfc/rfc8058.html). ### What should an agent be allowed to send? Promotional campaigns and automations on the warmed marketing identity. Transactional sends through a published automation on a transactional-purpose domain, fired with `POST /v1/automations/triggers/{triggerEventId}/fire`. Do not let it pick From addresses "to see what lands." ### Why review the first unattended send? Fluent copy will cite numbers you never published, or wear a receipt-shaped subject on a sale. Read it against the brief, seed-test Gmail, Yahoo, and Outlook, then confirm the headers.