AI Email Auto-Responder Using Ollama Local LLM

Overview

This workflow automatically monitors an IMAP email inbox every 5 minutes, reads unread emails, classifies them using a local Ollama LLM (Llama3), filters out spam, generates a professional draft reply, and prepares the output for review. It's a fully local AI-powered email assistant that helps you respond faster without sending data to external services.

Step-by-Step Walkthrough

1. Schedule Trigger (n8n-nodes-base.scheduleTrigger) — No auth. Runs the workflow every 5 minutes. Configured with an interval rule that triggers every 5 minutes.
2. Get Unread Emails (IMAP) (n8n-nodes-base.emailReadImap) — IMAP credentials (username, password, server). Fetches up to 10 unread emails from the configured IMAP inbox. The filter is set to readStatus: unread.
3. 1. Classify Email (Ollama) (n8n-nodes-base.httpRequest) — No auth (local API). Sends a POST request to http://localhost:11434/api/generate with a JSON body containing the model llama3:8b and a prompt that asks the LLM to classify the email into one of: URGENT, QUESTION, FEEDBACK, SPAM, OTHER. The response is parsed to extract the category.
4. Filter Spam (n8n-nodes-base.filter) — No auth. Checks if the classification result equals "SPAM" (case-insensitive). If it is SPAM, the email is dropped; otherwise, it continues to the next step.
5. 2. Draft Reply (Ollama) (n8n-nodes-base.httpRequest) — No auth (local API). Another POST request to the same Ollama endpoint, this time with a prompt that asks the LLM to write a professional, helpful reply. The prompt includes the original email's from, subject, and body (truncated to 2000 chars). The LLM generates the reply body.
6. Prepare Draft for Review (n8n-nodes-base.set) — No auth. Assembles the final output object with fields: draft_reply (the generated reply), original_from, original_subject (prefixed with "Re: "), and category (the classification). This output can be used to send a notification, store in a database, or manually review.

Setup Instructions

1. IMAP Email Account — You need an email account that supports IMAP (e.g., Gmail, Outlook, Yahoo). Enable IMAP access in your email settings and generate an app password if required. In n8n, create an IMAP credential with your email address, password, and server details (e.g., imap.gmail.com for Gmail).
2. Ollama — Install Ollama on your local machine or a server accessible from n8n. Pull the llama3:8b model (or any other model you prefer). Ensure the Ollama API is running on http://localhost:11434 (or update the URL in the HTTP Request nodes if running elsewhere).
3. n8n — Import this workflow into your n8n instance. Configure the IMAP credential in the "Get Unread Emails (IMAP)" node. If your Ollama instance is not on localhost, update the URL in both HTTP Request nodes. Save and activate the workflow.

Use Cases & Variations

• Customer Support — Automatically classify and draft replies to common support tickets. Add a human approval step before sending.
• Personal Email Assistant — Use it to quickly respond to non-urgent emails while you're away.
• Feedback Collection — Route FEEDBACK emails to a separate Slack channel or database.
• Model Customization — Swap llama3:8b for a smaller/faster model like mistral or a fine-tuned model for better accuracy.
• Auto-Send — Replace the final "Prepare Draft for Review" node with an SMTP node to automatically send the reply (use with caution).

{
  "name": "AI Email Auto-Responder (Ollama)",
  "nodes": [
    {
      "parameters": {
        "rule": {
          "interval": [
            {
              "field": "minutes",
              "minutesInterval": 5
            }
          ]
        }
      },
      "id": "schedule",
      "name": "Check Every 5 Minutes",
      "type": "n8n-nodes-base.scheduleTrigger",
      "typeVersion": 1.2,
      "position": [
        240,
        300
      ]
    },
    {
      "parameters": {
        "operation": "getAll",
        "returnAll": false,
        "limit": 10,
        "filters": {
          "readStatus": "unread"
        }
      },
      "id": "get-emails",
      "name": "Get Unread Emails (IMAP)",
      "type": "n8n-nodes-base.emailReadImap",
      "typeVersion": 2,
      "position": [
        460,
        300
      ],
      "credentials": {
        "imap": {
          "id": "REPLACE_WITH_YOUR_IMAP_CREDENTIAL_ID",
          "name": "Your IMAP Account"
        }
      }
    },
    {
      "parameters": {
        "url": "http://localhost:11434/api/generate",
// ... truncated (copy to see full JSON)