Inspiration
MailCheap is a free, open-source, self-hosted newsletter app — but self-hosting means wrestling with AWS SES, DNS records and spam rules, and setup is exactly where users quit. The support "team" is one founder (me, Pintu Kumar) and a written guide. I wanted to see whether an eGain AI Agent could carry that support load for real — not a tutorial bot, but an agent for a live product with a published knowledge base.
What it does
MailMate is MailCheap's AI support agent, built on the eGain AI Agent platform and trained on the real, published 9-step MailCheap setup guide at techbropro.xyz. It answers setup, configuration and cost questions from that guide with sources. A pre-chat workflow collects the visitor's name, email and topic, plus what they're stuck on, before the agent starts helping. Its distinctive feature is DNS Doctor: a process workflow that asks for the visitor's sending domain and runs a live diagnostic — querying public DNS (Google DNS-over-HTTPS API) for the domain's MAIL FROM records. It confirms the SPF record is live, or names the exact missing record and the value to add, pointing back to the guide. When MailMate can't solve a problem, an escalation workflow asks the urgency and emails the founder with the visitor's details and the full chat history, so the user never repeats themselves. Out-of-scope questions get an honest answer about what MailMate can help with — nothing is invented.
How I built it
Entirely on the eGain AI Agent platform (no-code Agentic Studio): pointed the agent at the live guide (website crawl plus the guide's Markdown version), let training derive intents (DNS and Domain Management, SES configuration, Cost Analysis, and more), then built three workflows on the canvas: (1) a pre-chat intake ending in Send to AI Agent with session context; (2) DNS Doctor as a process tool attached to the DNS intent — Question → API node (Google DNS-over-HTTPS) → Condition on the live response → outcome statements, with On Error and On Timeout handles wired to fallback messages; (3) an escalation workflow wired to both direct and knowledge-gap escalation, using Send Email with templated context variables and chat history. The agent is published via a public demo link layered over the guide site itself.
Challenges I ran into
eGain's connector actions (Send Email, Chat Transfer) are terminal nodes, so an "email then transfer" chain can't be built on one canvas — I redesigned the urgent path around a priority email with full context. The DNS Doctor's second planned check (MX) was dropped to keep the flow readable; the SPF/MAIL FROM check is the one that decides whether mail sends. I also verified obsessively: every workflow passed publish validation, and the full test matrix was run live — DNS Doctor pass branch on my real domain (mail.pintukr.in), missing-record branch on example.org, and a real escalation email received in my inbox.
Accomplishments that I'm proud of
A support agent for a real product anyone can try right now, whose centrepiece isn't a canned answer — it performs a real, live network diagnostic on the visitor's own domain during the conversation, and its escalation path was proven end-to-end with an actual email delivery.
What I learned
How much of support quality is workflow design rather than model answers: collecting context before the first answer, being honest at knowledge gaps, and making the handoff carry the whole conversation.
What's next for MailMate
Add the MX and DKIM checks to DNS Doctor, cover the SES production-access walkthrough as a guided process, and embed the agent on the MailCheap app itself.
Built With
- amazon-ses
- dns-over-https
- egain-agentic-studio
- egain-ai-agent
- google-public-dns
- mailcheap
