Skip to content

Email

Two kinds of mail

Neutron sends two quite different things, and most confusion about email here comes from mixing them up.

Ordinary mail. You write to someone, they write back, the whole exchange sits in a conversation you can read. Same as any mail client, except an agent can be in the conversation too.

Campaign mail. A message that goes out as part of an outbound campaign, under all the conditions described in Running a campaign — warmed domain, approved evidence, your explicit approval on each one.

They share the same plumbing: the same addresses, the same domains, the same warm-up rules. This page is that plumbing.

Receiving

Mail sent to an address Neutron manages arrives as a conversation, threaded by the person who wrote it. If the address belongs to a specific person or agent, it lands with them; the address decides who it reaches, which is why addresses are set up per person rather than per campaign.

One special case is handled without anybody looking at it. Every campaign message carries an unsubscribe address on its own domain, and anything arriving there suppresses that sender permanently, across every campaign. So does a reply whose entire subject or entire body is a bare unsubscribe, opt out, stop, or remove me. The match has to be the whole field — “please unsubscribe me from this one but keep the other” is not treated as an unsubscribe, because it is a sentence to a human, not a command.

Everything else a reply might mean — interested, not now, out of office, a question — is left for a person to read. That is deliberate. Guessing wrong in one direction suppresses a live prospect forever; guessing wrong in the other invents a deal nobody offered.

Writing a new email

Press + in the main navigation and choose New email. It opens a composer with four fields.

To. Starts suggesting from your contacts as you type, but accepts any address you type in full. You are not limited to people already in the system.

From. A dropdown of every address you are allowed to send as. See below — this is the field worth understanding.

Subject and body. Ordinary. There are generous length limits and the composer stops you before the server has to.

When it sends, you are taken to the conversation it created, so the sent message is read back from your mail rather than shown as a local echo. If it sends but cannot be saved into a conversation, you are told exactly that, and told not to send it again.

The From picker

Two kinds of address can appear in that dropdown, and they behave differently.

WHAT YOU PICK WHAT LANDS IN THEIR INBOX

The instance address [email protected] always offered · belongs to nobody the bare address, no name in front of it no signature is added A sending identity [email protected] belongs to one person or agent Nayeem Syed <[email protected]> that person's signature appended
The instance address is the one every message used before identities existed. It works, and it looks like a system address, because that is what it is.

The server decides what goes in that list, not the composer. You are offered your own addresses, plus the addresses of any agent you are allowed to send on behalf of, plus the instance address — and nothing else. Picking one you were not offered is not possible; if a composer was opened to answer as an address you are not entitled to, it names that address and refuses rather than quietly swapping in a different sender.

Replying

Replying inside a conversation defaults to the address the thread already speaks as — the address the other person wrote to. This is almost always what you want, and it means a thread does not change sender halfway through a conversation. You can override it with the same picker.

On a campaign’s Mail tab, the reply button names the address it will answer as, so you can see it before you click. If the campaign has no active sending address, the button is disabled and says so instead of failing when you press it.

Sending addresses (identities)

An identity is one address that one person or one agent is allowed to send as. You create it on their profile, not on a campaign and not on a domain — an address belongs to whoever sends from it.

Open the person’s or agent’s profile, find the identities panel, and press Add identity. You give it:

  • the part before the @ — you type this;
  • the domain — you pick this from a dropdown of domains Neutron manages, because an address on a domain nobody has set up cannot send;
  • a display name — the name that appears in the recipient’s inbox.

Afterwards, on the same panel, you can edit the display name, write a signature, and switch the address on or off with a toggle.

What the signature does. It is appended to the bottom of every message sent from that address, separated by the standard signature marker. That marker is not decoration — it is what mail clients look for when they trim quoted text, so your signature does not pile up at the bottom of a long thread. Only identities carry signatures; the instance address never does.

Switching an address off stops it sending immediately. Any campaign message queued against it will be held with “no active sender on this campaign” rather than going out.

If the panel says there is no managed sending domain, no domain has been set up on this instance yet, so there is nothing to hang an address on. Domains come first, addresses second.

Sending domains

A domain is set up once and then has to earn the right to send at volume. You will find them under Domains in the main navigation — one card per domain, showing where it is, which records are published, what its mail is doing to its reputation, and the button that stops it.

DNS pending DNS PENDING · 1/3 Verifying VERIFYING · bounce feed Warming up WARM-UP 3/7 · 40/day Healthy HEALTHY · 150/day

all three records published a real bounce report arrives seven clean days, one per stage

warm-up caps, one stage per clean day: 10 · 20 · 30 · 40 · 50 · 75 · 100 a day, then 150

Paused queued mail is held too many bounces, or you press pause
The two waits at the start are different things. DNS pending waits on you to publish records; verifying waits on the outside world to send back a real bounce report, proving the reputation feed works at all.

The states, in order

DNS pending. Three records must be published for the domain: DKIM, SPF, and DMARC. The card ticks each as it appears and shows a count like 1/3. Press recheck after you publish one. The DMARC tick also names the policy, because p=none is a published record that asks receiving servers to do nothing about a forgery — real, but weaker than it looks.

Verifying. The records are up, and the domain is now waiting for a genuine bounce or complaint report to arrive from the outside world. This is not busywork. A domain whose reputation feed is silently broken looks identical to a domain with a flawless record, and the second one is a lie you would act on. Nothing sends at volume until at least one real report has proven the feed is connected.

Warming up. The domain sends, at a cap that rises one stage per clean day: 10, then 20, 30, 40, 50, 75, 100 a day. The card shows this as WARM-UP 3/7. Stages are never skipped. A message that runs into the cap is not lost — it is held with “domain hit its daily cap” and goes the next day.

Healthy. Past the seventh stage, at a flat 150 a day.

Paused. Nothing sends. Every message already queued or approved on that domain is immediately held, saying “sending domain is paused”. Two things pause a domain: you pressing pause, and the watchdog.

The watchdog

A background check runs every six hours and judges each domain at most once a day.

It flags a domain when more than 5% of its mail bounces, or more than one message in a thousand draws a complaint. It pauses the domain when more than 10% bounces, or more than one in two hundred complains. Below 20 messages in a day it does not compute a rate at all and records the day as unmeasured — three bounces out of five sends is not a 60% bounce rate, it is not enough information.

It also pauses a domain that goes three days of real sending volume with no bounce reports at all. Perfect is not a plausible result at volume; it reads as a broken feed, and a broken feed is the failure this whole arrangement exists to catch.

Pressing resume on a paused domain starts it sending again. If whatever tripped the watchdog is still true, it will be paused again on the next pass — the button is honest about this in its tooltip. Resume after you have fixed the cause, not instead of.

Adding a new domain has no screen. The Domains view lists, rechecks, pauses and resumes domains — but there is no “add domain” form anywhere in the interface, and no way to create one from the outside. A new sending domain has to be put in place directly by whoever runs the instance. Everything after that point, from DNS checks through warming to the watchdog, is fully automatic and fully working.

Where things actually stand

Today’s state, 30 August 2026.

One sending domain: deepbrain.space, at warm-up stage 0 of 7. That means its cap is 10 messages a day and it has six more clean days of sending ahead of it before it reaches full rate. There is nothing to fix; it is doing exactly what it should.

Its sending address is [email protected]. The instance’s own address — the one that shows in the From picker as "(this instance)" — is [email protected].

Because no campaign can queue messages yet (see Running a campaign), the domain is warming without campaign traffic. Ordinary mail you send by hand from that address counts toward the same daily cap and the same reputation.

Where to go next

Last updated on