DEV Community: FloydBennett

I Built an Automated Real Estate Price Monitor with n8n + Octoparse MCP + Claude

FloydBennett — Thu, 11 Jun 2026 09:15:04 +0000

I've been casually watching the Seattle housing market for a while.
Every few days I'd open Zillow, filter by location, scroll through
listings, and manually copy prices into a spreadsheet. It was tedious
and I kept missing good deals.

One afternoon I thought — why not just automate this? A few hours
later I had a workflow that runs every morning at 8am, scrapes 800+
listings, saves everything to Google Sheets, and emails me a summary.
I haven't opened Zillow manually since.

Here's exactly how I built it.

TL;DR — I built a fully automated real estate price monitoring system for Seattle using n8n, Octoparse MCP, Claude AI, Google Sheets, and Gmail. It runs every morning at 8am, scrapes 800+ Zillow listings, saves them to a spreadsheet, and emails me a daily report — zero manual work required. Here's the exact step-by-step.

The Problem

Tracking real estate prices manually is painful. You open Zillow, filter by location, scroll through hundreds of listings, copy data into a spreadsheet, and repeat the next day. By the time you notice a price drop, someone else has already made an offer.

I wanted a system that does all of this automatically — scrape the data, store it, and send me a daily digest. No clicking. No copy-pasting. Just results in my inbox every morning.

Here's what I built.

What It Does

Every morning at 8am, this automated workflow:

Triggers on a schedule (n8n Schedule Trigger)
Scrapes 800+ Seattle home listings from Zillow via Octoparse MCP
Saves all data (address, price, beds, baths, sqft, status) to Google Sheets
Emails a daily summary report via Gmail

The entire pipeline runs without any human input. Once set up, it just works.

Tools Used

n8n — workflow automation (free, self-hosted or cloud)
Octoparse MCP — AI-powered web scraping via Model Context Protocol
Claude (via Anthropic API) — the AI agent that orchestrates the scraping
Google Sheets — data storage
Gmail — daily report delivery

Step 1 — Connect Octoparse MCP to Claude

The first step is authorizing Claude to use Octoparse MCP. This gives Claude the ability to search Octoparse templates, launch scraping tasks, and export data — all through natural language.

Once authorized, Claude has access to Octoparse's full template library and can run cloud scraping tasks on demand.

Step 2 — Test the Scraping in Claude

Before building the automation, I tested the scraping directly in Claude to make sure it worked. I asked Claude to search for Zillow templates on Octoparse:

Then I asked Claude to scrape Seattle real estate listings from Zillow:

"I'm building a real estate price monitoring system. Can you scrape home listings from Zillow using Octoparse? I'd like to track properties for sale in Seattle — please set the location to 'Seattle' and the type to 'buy'. I need up to 200 results including price, address, number of bedrooms, bathrooms, square footage, and listing status."

The task ran and collected 818 total listings. I asked Claude to export the first 20 results as a table:

The data was clean, structured, and ready to automate.

Step 3 — Set Up n8n

Now it was time to build the automation in n8n. I opened n8n and created a new workflow named "Seattle Real Estate Price Monitor".

Step 4 — Add a Schedule Trigger

The first node is a Schedule Trigger — this fires the workflow automatically every day at 8am.

Click "Add first step" → "On a schedule"

Configure the trigger:

Trigger Interval → Days
Days Between Triggers → 1
Trigger at Hour → 8am
Trigger at Minute → 0

Then click "Test this trigger" to verify it fires correctly.

Step 5 — Add an AI Agent Node

Click the "+" next to the Schedule Trigger → search "AI Agent" → select it.

In the AI Agent configuration:

Change Source for Prompt to "Define below"
Paste this fixed instruction into the Prompt field:

Scrape the latest homes for sale in Seattle from Zillow using
Octoparse. Use the Zillow Listing Scraper by keyword template, set
location to 'Seattle' and type to 'buy'. Get up to 50 results and
return the data including price, address, bedrooms, bathrooms, square
footage and status.

Step 6 — Configure the Chat Model

Scroll to the bottom of the AI Agent node. You'll see three sub-nodes: Chat Model, Memory, and Tool.

Click "+" under Chat Model → select "Anthropic Chat Model"

Set up your Anthropic API credential:

Go to console.anthropic.com
Click API Keys → Create Key → name it "n8n"
Copy the key and paste it into n8n

Note: If you prefer a free alternative, you can use Groq (console.groq.com) — create a free API key, select "OpenAI Chat Model" in n8n, set the Base URL to https://api.groq.com/openai/v1, and choose model llama-3.3-70b-versatile. Make sure to disable the "Use Responses API" toggle.

Heads up: I originally tried using the Anthropic API directly
but hit a rate limit issue on my first test run. Switched to Groq
as a fallback and it worked perfectly — and it's completely free.
If you're just testing, start with Groq.

Step 7 — Connect Octoparse MCP as a Tool

Click "+" under Tool → the MCP servers panel appears.

Octoparse isn't in the default list — that's fine. Click "MCP Client Tool" at the top.

Configure the MCP Client:

Endpoint → https://mcp.octoparse.com
Server Transport → HTTP Streamable
Authentication → Header Auth

Click "Set up credential" and fill in:

Name → Authorization
Value → Bearer YOUR_OCTOPARSE_API_KEY

Get your Octoparse API key at octoparse.com/console/account-center/api-keys

Click "Execute step" to test the connection. If successful, you'll see the Octoparse tool list:

MCP connection successful! The AI Agent now has full access to Octoparse's scraping capabilities.

Step 8 — Add Google Sheets to Store the Data

Click "+" on the right side of the AI Agent node → "Action in an app" → search "Google Sheets" → select "Append row in sheet"

Click "Sign in with Google" to authorize. Once connected, you'll see:

Now create a Google Sheet:

Go to sheets.google.com and create a new spreadsheet
Name it "Seattle Real Estate Price Monitor"
Add these column headers in row 1: Date | Address | Price | Beds | Baths | SqFt | Status

Back in n8n, select your spreadsheet in the Document dropdown and Sheet1 in the Sheet dropdown.

Map the fields:

Date → {{ $now.toISO() }}
Address / Price / Beds / Baths / SqFt / Status → {{ $json.output }}

Step 9 — Add Gmail to Send the Daily Report

Click "+" on the right of the Google Sheets node → search "Gmail" → select "Send a message"

Click "Sign in with Google" to authorize Gmail.

Fill in:

To → your email address
Subject → Daily Seattle Real Estate Price Report - {{ $now.format('yyyy-MM-dd') }}
Email Type → HTML
Message → {{ $json.output }}

Step 10 — The Complete Workflow

Here's the full workflow with all nodes connected:

Schedule Trigger → AI Agent → Append row in sheet → Send a message
                      ↓
          Anthropic Chat Model + MCP Client (Octoparse)

Step 11 — Execute and Verify

Click "Execute workflow" to run a full test.

The AI Agent calls Octoparse MCP, which scrapes Zillow, finds the right template, and returns results.

818 listings collected. The Google Sheets file updated automatically:

And the Gmail report arrived in my inbox:

Workflow executed successfully.

One thing I didn't expect — the Zillow scrape takes about 4-5 minutes
to collect 800+ listings. The first time I ran it I thought something
had broken, but Claude was just patiently polling for results. Worth
knowing before your first test run.

Results

818 Seattle home listings scraped from Zillow in a single run
Data saved to Google Sheets with timestamp, address, price, beds, baths, sqft, and status
Daily email report delivered automatically every morning at 8am
Zero manual work after initial setup

Why Octoparse MCP?

Traditional web scraping requires writing custom code for each website, maintaining scrapers when sites change, and managing proxies and rate limits. Octoparse MCP eliminates all of that.

You describe what you want in plain English. Claude finds the right template, configures it with your parameters, runs the task, and returns structured data. It's the difference between writing a web scraper and asking someone to do it for you.

Combined with n8n for scheduling and Google Sheets for storage, this becomes a complete data pipeline — built in an afternoon, runs forever.

Adapting This for Other Use Cases

The same pattern works for any recurring data monitoring task:

Job listings — track new openings at target companies daily
Product prices — monitor competitor pricing on Amazon or e-commerce sites
News monitoring — scrape industry news and send daily digests
Lead generation — collect business listings from Google Maps on a schedule

Change the Octoparse template and the prompt in the AI Agent node — everything else stays the same.

Get Started

Octoparse MCP: octoparse.com/mcp
n8n: n8n.io (free tier available)
Octoparse template library: octoparse.com/template

Built for the Octoparse MCP Challenge · Track 1: Build a Workflow with Octoparse MCP

Powered by Octoparse MCP + n8n + Claude

I Built an AI-Powered World Cup News Digest Page in Under a Minute — Using Octoparse MCP + Claude

FloydBennett — Thu, 11 Jun 2026 08:44:01 +0000

TL;DR — I built an Octoparse Agent Skill that scrapes news, summarizes it with Claude, and outputs a single deployable HTML page with live scores, market odds, fan sentiment, and more. The World Cup 2026 demo took under a minute to generate. Here's exactly how it works.

The Problem with News Aggregators

Most news tools give you a list of links. You still have to click every article, read through the noise, and figure out what actually matters.

I wanted something different: tell Claude a topic, get back a finished product — a page I could open in my browser, share with friends, or deploy to a URL in two minutes.

That's what this Octoparse Agent Skill does.

What It Produces

Here's the World Cup 2026 Matchday 01 example — generated on June 11, 2026, the opening day of the tournament:

The page contains seven modules:

AI-Generated Headline — Claude distills all scraped articles into one punchy 2–3 line statement
Live Scoreboard — real-time match scores, kickoff times, today's fixtures
Group Standings — all 12 World Cup groups
Prediction Wall — market odds scraped via Octoparse MCP, visualized as probability bars
Fan Sentiment Radar — Reddit + YouTube comment sentiment for all 4 teams playing today
News Feed — 18 articles grouped by theme, each with a "Why it matters" AI insight
Alert Section — high-priority controversy and political stories flagged in red
Visual Field Map — 48-team grid with today's active squads highlighted
Share Modal — cyberpunk-styled card with one-click sharing to X/WhatsApp

Everything is in a single self-contained HTML file (~420KB) with fonts embedded. No backend, no server, no dependencies.

Live Demo

👉 ohayo716.github.io/octoparse-news-digest/worldcup-digest.html

How It Works — The Full Pipeline

User types topic (or drags SKILL.md and says "surprise me")
      ↓
Claude reads SKILL.md instructions
      ↓
Octoparse MCP finds matching scraping template
      ↓
Octoparse scrapes news + odds + sentiment data
      ↓
Claude summarizes articles + generates AI headline
      ↓
Live score API fetches real-time match data
      ↓
Single HTML file generated and delivered
      ↓
User opens in browser or deploys to GitHub Pages

The entire pipeline runs inside a Claude conversation. The user never touches code.

The Skill File — SKILL.md

The core of this project is a single SKILL.md file. This is an Octoparse Agent Skill — a markdown file that gives Claude structured instructions for using Octoparse MCP.

Step 0 — Drag and Drop Trigger

When a user drags SKILL.md into Claude Desktop, the skill fires immediately:

"I'll build you an AI news digest page right now. What topic should I cover? Or say 'surprise me' and I'll pick today's biggest story."

No configuration needed. No setup. Just drag and go.

Step 1 — Topic Detection

The skill auto-detects context. During the World Cup, it automatically recognizes the tournament as the dominant topic. The user can override with any keyword — "AI regulation", "fintech", "Tesla", anything.

Step 2 — One Optional Question

Before scraping, the skill asks once whether the user has a free API key for live data. If yes, it embeds it directly in the HTML output. If no, the page works perfectly without it. No file editing required.

Step 3–4 — Octoparse MCP Scraping

The skill searches Octoparse for matching templates:

1. Google News [topic]       → broadest coverage
2. [Site name] articles      → if user named a source  
3. Odds/betting scrapers     → for prediction wall
4. Reddit/YouTube scrapers   → for sentiment analysis

Step 5 — Claude Summarization + AI Headline

For each article, Claude generates:

What happened (1 sentence)
Why it matters (1 sentence, specific and insightful)

Then produces the AI Headline — distilled from everything scraped:

"THE AZTECA ROARS. THE WORLD STOPS. FOOTBALL IS BACK."

Step 6 — HTML Generation

Claude outputs a single self-contained HTML file. Fonts are embedded as base64 so the file works completely offline once generated.

The Design Language

The aesthetic is cyberpunk terminal — deep dark backgrounds, fluorescent green accents, scanline overlay, monospace fonts. It signals immediately: this was generated by a machine, and that's the point.

Three fonts, each with a specific role:

Font	Used for	Why
BigShoulders Bold	Hero titles, score digits	Maximum visual impact
Silkscreen	Section numbers `01 02 03` and labels only	Pixel aesthetic, navigation feel
GeistMono	All body text, news titles, metadata	Clean, readable at any size

The section dividers use Silkscreen pixel font — everything else uses GeistMono. Pixel aesthetic present without sacrificing readability.

Prediction Wall — Market Odds via Octoparse

The Prediction Wall shows market odds for today's matches, scraped via Octoparse MCP from public betting aggregators and visualized as probability bars.

Instead of users voting on predictions, the market does the predicting. Odds are the most accurate real-time signal of what experts think will happen — and Octoparse can scrape them automatically.

Fan Sentiment Radar — Reddit + YouTube via Octoparse

The sentiment module scrapes Reddit threads and YouTube comments for all four teams playing today, then Claude analyzes the text for positive/neutral/negative sentiment and extracts the most-discussed keywords.

Each team gets:

Three sentiment bars (positive / neutral / negative %)
8 hot keywords from fan discussions
Post count and source attribution

This turns raw social data into something instantly readable.

News Feed — AI Summarized

Every article card follows the same structure:

[SOURCE · TIME AGO]              [TAG]
[News headline — GeistMono, readable]
┌─ // WHY IT MATTERS ──────────────┐
│  [1-sentence AI insight]         │
└──────────────────────────────────┘
> READ FULL ARTICLE →  [real URL]

Articles are grouped into themes: On the Pitch, Off the Pitch, Alert (controversy).

Visual Field Map

A 48-team grid where today's active squads pulse in fluorescent green. One glance tells you who's playing today out of all 48 nations.

The Share Modal

Clicking "⚡ SHARE SIGNAL" opens a cyberpunk-styled modal with:

Data-flow animation (green light streaming down the right edge)
Scanline overlay effect
AI headline displayed prominently
Today's match and stats
One-click sharing to X (Twitter) or WhatsApp
Copy-to-clipboard for any platform

The card is designed to look stunning as a screenshot — the kind of thing someone would actually post.

System Log Footer

Every generated page ends with a system log showing the full pipeline:

2026-06-11 00:00:01  [OK]  Octoparse MCP connected · templates loaded
2026-06-11 00:00:04  [OK]  Scrape executed · 18 articles · 6 sources
2026-06-11 00:00:08  [OK]  Claude summarization complete · 3 themes identified
2026-06-11 00:00:09  [OK]  AI headline generated · distilled from 18 articles
2026-06-11 00:00:11  [OK]  HTML rendered · single file · fonts embedded · ready
2026-06-11 00:00:12  [OK]  Deploy → GitHub Pages · done

Transparency as design — the user sees exactly what Octoparse and Claude did.

Deploy to GitHub Pages (2 minutes)

Create a new GitHub repository
Upload worldcup-digest.html, rename it index.html
Go to Settings → Pages → Source: main branch
Live at https://[username].github.io/[repo-name]/

Once deployed, live APIs work correctly (local file CORS restrictions don't apply).

Use It for Any Topic

The World Cup is just the demo. The skill works for any topic:

"Make me a news digest about AI regulation"
→ Scrapes Reuters, TechCrunch, The Verge
→ Groups by: EU AI Act / US Congress / Corporate compliance
→ Same cyberpunk digest page, different data

"Build a fintech briefing for this week"  
→ Scrapes Bloomberg, FT, CoinDesk
→ Groups by: Funding / Regulation / Market moves

"Surprise me"
→ Claude detects today's biggest story automatically
→ Builds the digest without any further questions

Get the Skill

Everything is open source:

📁 GitHub: github.com/ohayo716/octoparse-news-digest
🌐 Live demo: ohayo716.github.io/octoparse-news-digest/worldcup-digest.html

To use it:

Download SKILL.md from the repo
Connect Octoparse MCP to Claude Desktop (setup guide)
Drag SKILL.md into a Claude conversation
Say your topic — or "surprise me"

What's Next

Match event timeline — goals, red cards, VAR decisions as a visual timeline
Multi-language output — French and Chinese versions of the digest
n8n automation — scheduled daily runs, auto-deploy to GitHub Pages
More topics — AI news, fintech, startup funding rounds

If you build something with this skill, drop a link in the comments — I'd love to see it.

Built for the Octoparse MCP Challenge · Track 2: Build an Octoparse Skill