AI Email App: Build Your Own Smart Inbox with n8n and an LLM
An AI email app you build yourself: n8n workflows that summarize, triage, and auto-reply with your chosen LLM. Full recipe included.
💡 What You Will Learn
An AI email app you build yourself: n8n workflows that summarize, triage, and auto-reply with your chosen LLM. Full recipe included.
📜 Table of Contents
You do not need to wait for a startup to make your inbox smart. With n8n and any LLM, you can build an AI email app that summarizes every thread, flags what matters, and drafts replies - wired to your exact workflow.
The Recipe
Trigger: n8n (199,535 stars) Gmail trigger on new mail. Step 2: an LLM node summarizes the thread and scores importance 1-5. Step 3: rules route - score 4-5 goes to your inbox, 1-2 auto-archives with a label, 3 waits for your review. Step 4: for known reply patterns (meeting requests, price quotes), the LLM drafts a reply you approve in one click.
Model choice: use a cheap model for triage (fast, high volume) and a stronger one for drafts. Connect via LiteLLM (55,701 stars) to switch providers without touching the workflow. Privacy: self-host n8n and point at Ollama (177,902 stars) for a fully private pipeline.
Comparison
| Step | Action | Tool |
|---|---|---|
| 1 | Trigger on new mail | n8n Gmail |
| 2 | Summarize + score | LLM node |
| 3 | Route by score | n8n rules |
| 4 | Draft replies | LLM + approval |
FAQ
Q: How long does this take to build?
A: A basic triage workflow takes 1-2 hours in n8n; the draft-and-approve loop adds another hour.
Q: Which model should I use?
A: Cheap model (DeepSeek-class) for triage, stronger model for drafts. Route both through one gateway.
Written by our editorial team; tools listed here are tested or verified against public sources. Links point to official sites or GitHub repos for reference only โ no paid placements.