B
Claude Code logoClaude CodeRun agent workflows with zero data leaving your network

Fully Local No-Cloud Automation Agent (n8n + Ollama)

setuproll@setuproll
81.0Overall score

A self-hosted n8n plus Ollama setup where an open model drives the reasoning, so sensitive files and customer data never touch a third-party API. For privacy-strict teams, regulated industries and homelab owners who want automation without sending anything to the cloud.

81.0Score
960Votes
5Components

Install this build

Export
terminal
ollama serve & docker run -p 5678:5678 -e OLLAMA_HOST=host.docker.internal n8nio/n8n

Components

Model

  • Qwen3 32B (Q4_K_M)
  • Llama 3.3 70B on 64GB+ boxes

Stack

  • n8n (self-hosted)
  • Ollama OpenAI-compatible endpoint
  • Qdrant

Tools

  • Local filesystem node
  • Qdrant retrieval
  • SMTP for internal mail

Trigger

  • New file in a watched folder, or schedule

How it works

  • n8n and Ollama run on the same box behind your firewall
  • A watched-folder or cron trigger starts the workflow
  • n8n calls Ollama via its OpenAI-compatible /v1 endpoint
  • The model summarizes, classifies or extracts, grounded by Qdrant
  • Results stay internal: written to disk, DB or local email only

Summary

A self-hosted n8n plus Ollama setup where an open model drives the reasoning, so sensitive files and customer data never touch a third-party API. For privacy-strict teams, regulated industries and homelab owners who want automation without sending anything to the cloud.

81.0 score 960 votes

0 Reviews

Your rating
Sign in to post

Loading discussion...

← All builds