Skip to main content
aiAdaptiv
All case studies

Reference build — our own collection, published

Seventy-six automations, yours to take

Most automation pitches show you a five-node demo. This is the opposite: the workflows we actually built and ran, exported exactly as they are and put in a public repository. Import the JSON, plug in your own credentials, keep whatever is useful.

  • Client

    None — our own collection

  • Field

    Automations and agents, built in n8n

  • Published

    Public GitHub repository, MIT

  • Format

    76 importable workflow JSON files

The export is the cheap part. Deciding which automation is worth running, and keeping it alive when the API changes, is the work.

  • 76workflows in the repository1,061 nodes in total — 14 per workflow on average
  • 57of them call a language model39 contain an agent, 24 query a vector store
  • 19third-party services wired upTelegram, Gmail, Notion, Drive, Sheets, Postgres, Slack, Supabase, Calendar…
  • MITlicensed, so you can actually use themcommercially too, inside your own product, no attribution required

The problem: the demo is not the work

Automation demos are five nodes long because five nodes fit in a screenshot. Real ones are not. The average workflow in this collection is fourteen nodes, and the extra nine are the parts nobody films: the branch for when the API answers with nothing, the batching so a thousand rows do not become a thousand calls, the retry, the place the failure gets reported.

That gap is why so many automations are built once, impress everybody, and are quietly switched off a month later.

What is in the repository

Everything from small utilities to full agents, grouped roughly by what they do. Seventeen of them were switched on at the moment of the backup — these are working automations, exported as they ran, not tidied-up examples written for an article.

  • Assistants with a job: inbox, calendar, contacts, invoices, travel, research, a personal trainer
  • RAG knowledge bases over Notion, Drive, newsletters and YouTube transcripts, on Postgres with pgvector
  • Collectors: Google Maps, X, competitor ad libraries, web scraping with Firecrawl
  • Content pipelines: transcript to summary, research to draft, SEO agents working as a team
  • Image and video generation wired into a Telegram front end, so you use it from your phone

Why we give them away

Because the export is the cheap part. Anybody can import a JSON file; the difficulty starts immediately afterwards, and none of it is in the file.

Which of these is worth running for your business, and which will cost more attention than it saves. What it costs per month once real volume goes through it. Whose credentials it holds and what happens when that person leaves. Who notices when it breaks at two in the morning. And keeping it alive when a provider changes an endpoint, which they do.

A collection you can read is a better argument than a page of adjectives about what we could build. Take it, run it, and if the second part is the part you do not want to own, that is the conversation to have.

Go and look

The whole collection is public and MIT licensed. The README indexes every workflow by name with the services it touches, because the file names are n8n's own ids. Every number on this page comes from parsing those files, so you can recount them yourself — they are a snapshot of our n8n instance rather than a maintained product, so expect to update node versions and to bring your own credentials.

github.com/Kacper-Hernacki/n8n-templates-backup

What this case does not claim

This is our own collection, not a client engagement, and nobody has audited it. We are not claiming hours saved or money made, by us or by anyone who downloads it — we have not measured that. The workflows are a backup taken in December 2025: some will need their node versions updated, several were built for a specific account and will need rewiring, and a few are frankly rough. They are published as they are, which is the only honest way to publish them.

The second part is the part we do

Choosing what is worth automating, running it where your data is allowed to be, and being the person who gets paged when it breaks. Bring a process that eats your week and we will tell you whether it is worth the wiring.