Why I Built a Second Brain in n8n — And How You Can Too

I spent years collecting notes, highlights, bookmarks, and ideas across a dozen different apps. My browser had 400+ tabs. My Notion had databases I hadn't opened in months. My Readwise was full of highlights I told myself I would review someday.

The core problem was not that I lacked information. It was that I lacked a system for information to surface at the right time.

A second brain is not a note-taking app. It is a workflow. And after trying Notion dashboards, Roam graphs, and Obsidian vaults, I ended up building mine in n8n — the open-source workflow automation platform. Here is why, and exactly how to build your own copy.

Why n8n instead of a notes app

Most second-brain systems are static. You write notes, tag them, maybe link them, and then you hope you remember to look at them later. That works if you are disciplined about reviews, but most of us are not.

n8n flips the model: instead of you pulling information into your brain, workflows push information at you when it matters.

What n8n gives you that a notes app cannot:

• Trigger-based capture — a Telegram message, an email forward, a saved tweet — all automatically ingested
• AI enrichment — incoming notes get summarized, categorized, and linked before they reach your database
• Scheduled review prompts — weekly digests of what you saved, surfaced in your messaging app of choice
• Cross-app piping — Readwise highlights → n8n → Notion → periodic Slack digest, all in one flow

The result is not another notebook. It is an active system that processes, connects, and resurfaces knowledge without you touching it.

How the system is structured

My second brain has three layers, each handled by a dedicated n8n workflow:

Layer 1: Capture (the inbox)

Every piece of information enters through one of four gates:

1. Telegram bot — I forward links, voice memos, or quick notes to a private Telegram channel. n8n watches the channel via a Telegram trigger.
2. Email alias — Newsletters I want to mine get forwarded to an alias. n8n's email trigger (IMAP) polls and extracts the body.
3. Twitter saved tweets — A periodic n8n workflow fetches my saved tweets and parses the text and links.
4. Browser bookmarklet — A tiny script sends the current page URL + selected text to a webhook that n8n listens on.

All four feeds land in a single Notion database called "Inbox." No tags, no folders, no decisions required at capture time. That is the most important rule: capture must be frictionless or you will stop doing it.

Layer 2: Processing (the AI enrichment)

Every 24 hours, a scheduled workflow runs through unprocessed Inbox items and does three things:

1. Summarize — sends the content to an OpenAI call with a prompt: "Summarize this in 2–3 sentences. What is the actionable takeaway?"
2. Categorize — a second prompt classifies the item: "Is this a: tool recommendation, workflow idea, content inspiration, business model insight, or reference?"
3. Link — searches existing Notion database entries for related topics using embedding similarity and appends a "related to" relation

The processed item then moves from "Inbox" to "Library" with all three AI-generated fields populated.

Layer 3: Resurfacing (the review loop)

A weekly workflow runs every Sunday evening and:

1. Pulls 5 random items from the Library that I have not reviewed in the past 14 days
2. Sends them as a Telegram message with the summary and a link back to the Notion entry
3. Adds a "resurfaced on" date stamp so the same item does not appear again soon

This is the layer most second-brain systems skip. Capturing is easy. Reviewing is the hard part. Automating the review loop is what makes the system actually useful.

The n8n workflow diagram

The full setup uses three connected workflows:

Workflow 1 — Telegram Inbox
  Trigger: Telegram message received
  Action: Create Notion database item (Inbox)
  Action: Send confirmation back to Telegram

Workflow 2 — Daily Processing
  Trigger: Schedule (daily, 2 AM)
  Action: Query Notion for unprocessed Inbox items
  Action: Loop through each item
    HTTP: Send content to OpenAI (summarize + categorize)
    Action: Update Notion item with AI fields
    Action: Move item to Library database
    HTTP: Compute embedding, store in Library

Workflow 3 — Weekly Review
  Trigger: Schedule (weekly, Sunday 6 PM)
  Action: Query Notion Library for items not reviewed in 14 days
  Action: Randomly select 5
  Action: Send Telegram message with summaries + links
  Action: Update "last reviewed" date

What this replaces

Before building this, I was paying for:

I self-host n8n on a $6/mo VPS, so the monthly savings are around $22, and I get more control over the actual logic.

What I would change if I rebuilt today

The system works well, but there are three things I would do differently:

1. Add a vector database earlier. Embedding search in Notion is workable but slow at scale. If I rebuilt today, I would pipe processed items into a Supabase + pgvector setup for instant semantic search across everything I have saved.

2. Use AI tagging more aggressively. My two-prompt approach (summarize + categorize) is good but misses nuance. A single prompt with structured output (JSON mode) would be cleaner and faster.

3. Build a web dashboard. Telegram as the front-end works, but a simple Next.js dashboard with a search bar and filterable library view would make the system accessible on desktop without scrolling through chat history.

How to build your own in a weekend

Here is the fastest path to a working second brain with n8n:

1. Set up n8n — spin up a free n8n.cloud account (2000 monthly workflow executions on the free tier) or self-host on Railway/Render for pocket change
2. Create a Telegram bot — talk to BotFather, get a token, and create a private channel for your captures
3. Build the inbox workflow — Telegram trigger → Notion "Inbox" item. This takes 15 minutes.
4. Add an OpenAI node — send new items to GPT-4o-mini for summarization. Costs pennies per thousand items.
5. Schedule the review — set a weekly Telegram message with random items from your library

You do not need all three layers on day one. Start with capture. Add enrichment when you have 50+ items. Add resurfacing when you notice you are not reviewing.

Why this matters for creators

A second brain is not just a productivity flex. For creators, it is a content engine.

Every week, the weekly review surfaces something I saved months ago that I had forgotten about. That becomes a newsletter topic, a Twitter thread, or a blog post draft. The system does not just store knowledge — it generates output.

If you publish anything regularly, your biggest bottleneck is not writing. It is finding what to write about. A well-built n8n second brain solves that better than any notes app I have tried.