Hermes reasons capably on an open model foundation, and the judgement this template needs is narrow but real: is this one dead address, or the first sign of a reputation problem. Several recipients at the same domain failing together points one way and a single typo points the other. The CRM write stays fenced to the fields your rules name.
Run a staged conversation — no account needed. The agent handles it for real while a simulated world answers its tool calls; nothing touches real accounts, and nothing is actually sent.
Maartje de Vries
550 5.1.1 The email account that you tried to reach does not exist. Please try double-checking the recipient's email address for typos. First permanent failure for this address, sent by the July invoice reminder campaign. The contact opened eleven emails before this and last clicked on 14 June.
Using this template drops you into a guided setup. It asks exactly this, nothing else:
Connect Mailgun
One sign-in. The agent acts through your account, scoped to what this template uses.
Connect HubSpot
One sign-in. The agent acts through your account, scoped to what this template uses.
Connect Slack
One sign-in. The agent acts through your account, scoped to what this template uses.
List hygiene rules
What a hard bounced address should become in HubSpot: which field to set, to which value, and where the bounce details get written. The agent only ever makes the changes this names.
Alert criteria
Which bounces are worth interrupting someone for, for example a paying customer domain, a VIP account, or more than five bounces on one domain in a day. Everything else stays silent.
Runs on Hermes
Preselected for this page — connect your Hermes account during setup, or switch to NoClick's built-in models with one click.
Watch it handle a test run
A staged conversation against a simulated world — then it’s live.
Recognising a domain-level failure means reading several bounces together rather than classifying one in isolation.
When it looks like a blocklisting, the correct move is to change nothing and hand a person the evidence. That restraint is the behaviour worth verifying.
The writes are constrained to fields you name, on a contact matched exactly by the bounced address, with a hard stop when zero or several contacts match. That containment makes the harness choice a smaller question than it first appears. Run Shared Inbox Rejection, where two contacts share one address and the correct answer is to change nothing.
Free to start. Guided setup, a test run against staged conversations, and it's live.