Meet Adamko, Lettr's AI assistant
Adamko is Lettr's built-in AI assistant. He sets up a new account from nothing but a website URL, creates branded templates and campaign drafts, diagnoses DNS and delivery problems from live account data, and records every change he makes in a team-wide activity log with one-click undo.
Adamko is Lettr's built-in AI assistant. He lives in a chat drawer in the dashboard and does real work in the account: ask him to set up a fresh account and he builds a brand kit from the company website, scaffolds branded transactional emails, and registers a sending domain; ask why a specific recipient never got an email and he pulls the delivery events and suppression records and names the likely cause. Every answer draws on live data from the account, and every change he makes lands in an activity log where it can be undone.
Not a search box over the docs
Adamko's answers are grounded in live data from the account: its domains, campaigns, event history, and suppression lists. The question "what's my bounce rate this week?" triggers a real query, and "why isn't my domain verifying?" triggers a DNS lookup at that moment, not a recitation of what SPF is. When the answer genuinely lives in the documentation, he searches it and cites the article.
Adamko is also Lettr's second AI surface, and the two are complementary. The MCP integration brings the Lettr API into an external assistant such as Claude or Cursor, which suits developers already working in one. Adamko lives inside the dashboard, needs no setup, and can do things an API call cannot, like highlighting the exact button a task needs next.
Account setup from a website URL
The flagship flow is onboarding. On a fresh account, Adamko offers to set everything up from a single input: the company's website URL. He visits the site and builds a brand kit from it (logo, brand colors, fonts, tone of voice, and the language the site is written in), then recommends a set of transactional emails that fits the business. A SaaS gets welcome, verification, password reset, and invoice emails pre-selected; an e-commerce store gets a different mix, and billing emails are never suggested to a site that doesn't charge anyone.
After one confirmation, he creates the selected templates in the brand design with a shared header and footer, registers a sending domain, and issues the first API key together with a working integration snippet for the chosen stack. The one thing he cannot do is edit DNS at the registrar. He shows the exact records to add, each value click-to-copy, and confirms them with a live check once they are in place.
Templates, campaigns, and a plan
Content creation runs off the same brand kit. Adamko scaffolds any of the eleven transactional email types in his catalog, builds custom templates from a plain description ("a template announcing our dark mode feature, button linking to the settings page"), and drafts whole campaigns: subject line, body copy in the brand voice, layout in the brand design. Everything he generates lands in the account as a normal, fully editable template or draft; there is no separate "AI content" silo.
Two smaller jobs round this out. Before a big send, he can review a template for broken merge tags, a missing unsubscribe link, layout risks, and spam triggers, each finding with a concrete fix. And "plan my campaigns for next quarter" produces a dated sending calendar grounded in the account's real audience sizes and remaining quota, drafted one campaign at a time as each is approved.
Diagnostics from evidence, not guesses
Diagnostics is where Adamko saves the most time, because a delivery problem usually means cross-referencing DNS records, event logs, suppression lists, and webhook configs. Adamko pulls from all of them at once. The classic support question, "why didn't jane@acme.com receive my email?", gets answered end to end: he checks the recipient's recent message events, the suppression list, and the domain state, and answers with one likely root cause, the evidence, and the fix.
The same applies at account level. "Check my sending health" returns a single verdict (healthy, needs attention, or at risk) backed by domain verification state, bounce and complaint trends, and suppression estimates. "Why are my API requests failing?" samples the recent request logs and webhook configuration and returns issues ordered by severity. On verification he is deliberately strict: a domain is only ever called verified after a live DNS check, never from memory.
He answers with a button
"Where do I configure X?" gets a button, not a paragraph. For multi-step jobs, replies carry a Guide me action that opens the right page and walks through the task one element at a time: a highlight rings the exact field, a bubble explains the step, and the tour advances on its own as each step completes.
He also watches for a small set of situations worth flagging (a domain unverified for more than a day, a bounce rate past the warning threshold, an abandoned campaign draft) and surfaces each as a dismissible banner with the diagnosis one click away. Nudges alert; they never act.
What he will never do
An assistant with write access needs hard limits, and Adamko's are deliberate. He will never:
- Send or schedule an email or campaign; sending is always a human click
- Create segments, lists, or topics on his own; he suggests them and hands off to the builder, since membership decides who gets emailed
- Invite teammates or change roles
- Reveal an API key; keys appear once at creation in a copy-once card that bypasses the model entirely
- Read a dropped file; a CSV goes straight to the importer, and Adamko sees only its structured analysis, never the raw rows
Everything he does do is recorded in a team-wide activity log: who asked, the exact prompt, and what was created, updated, or deleted. Most entries carry an Undo button, and the rule for it is not a time limit but "untouched since". A drafted campaign can be undone while it is still an unedited draft; a sending domain while it is unverified; a template while nobody has changed it. Once a person builds on top of his change, the button disappears, because one-click reversal would then destroy human work.
FAQ
Can Adamko send emails on his own?
What happens if Adamko creates something I didn't want?
Can Adamko change my DNS records for me?
Does Adamko work in languages other than English?
Does Adamko have usage limits?
Bottom line
Adamko is the fastest route from an empty Lettr account to a production-ready one, and after that, the first place to ask when something misbehaves. Trying him costs nothing: create a free Lettr account and ask him to set everything up. Once you're in, the AI assistant docs walk through everything he can do.