Email AI Assistant
5 min → 30 sec per email reply
5 min → 30 sec
per email reply
∞
regeneration loops until perfect
auto
detects reply language
The problem
Email takes 5 minutes per reply — not because the replies are hard, but because switching context, reading the thread, and matching the right tone all add up. At 20 emails a day, that's over an hour of overhead.
How it works
Email arrives
Gmail trigger fires and sends the full email to the AI agent
AI drafts a reply
GPT-4o reads the tone and writes a natural response in the same language
Preview lands in Telegram
You see the sender, tone, email content, and the planned reply — no inbox switching
Write 'ok' or a correction
One word to send it — or type any edit to improve the draft
AI rewrites instantly
Your correction becomes context — a new draft arrives in seconds
Reply sent
Email goes out, pending data clears, inbox stays clean
The payoff
5 min → 30 sec per reply
20 emails/day × 4.5 min saved = 90 min returned daily. That's 7.5 hours a week — nearly a full working day.
Auto language detection
Received in English? Replies in English. In Ukrainian? In Ukrainian. No extra configuration needed.
Infinite regeneration
Write any correction instead of 'ok' — the AI rewrites with full original context preserved. No limit on loops.
No accidental sends
Two-flow architecture makes it physically impossible to send without your explicit approval.
What's under the hood
n8n — two workflows
Gmail trigger flow and Telegram trigger flow connected via shared static data. Each runs independently.
GPT-4o via OpenRouter
Structured output parser extracts email fields reliably. Regeneration agent takes previous draft + correction as context.
Static data as state
n8n workflow static data stores the pending email between flows — no external database needed.
Free. No catch.
The workflow JSON is open — anyone can import it into their n8n instance and connect their own Gmail and Telegram accounts.
Email reading
Polls Gmail every minute, extracts sender, subject, tone, and full content automatically
AI draft
GPT-4o writes a natural reply — casual or formal depending on the tone of the original
One-tap approval
Reply preview lands in Telegram. Write 'ok' to send. You stay in control, nothing goes out without you.
Regeneration loop
Write a correction instead of 'ok' — the AI rewrites the reply and shows the new version. Loop until perfect.