Multi-Platform Review Scraper MCP

Scrape customer reviews, ratings, and sentiment analysis from Google, Trustpilot, Yelp, TripAdvisor, and Amazon in a single AI agent tool call. Returns individual review records with reviewer metadata, aggregate cross-platform ratings, sentiment classification, and fake review detection.

Overview

This MCP server gives AI agents live access to customer review data across five major platforms. For any business URL, it returns:

• Individual review records — star rating, title, text, author name, date, verified badge, helpful votes
• Cross-platform summary — aggregate average rating, total review counts per platform, sentiment breakdown
• Fake review detection — probability scoring and flagging of suspicious reviews
• Export functionality — JSON, CSV, or Markdown export of review data

Review data is extracted from public HTML using structured data selectors and JSON-LD schemas. No API keys required for any platform.

Tools

scrapeReviews

Scrape individual customer reviews and aggregate summary from one or all platforms.

Parameters:

Parameter	Type	Required	Description
businessUrl	string	Yes	Business URL on a review platform
businessName	string	No	Business name to disambiguate searches
location	string	No	City/address to disambiguate multi-location businesses
platform	string	No	Target platform: google, trustpilot, yelp, tripadvisor, amazon, or all (default: all)
maxReviews	integer	No	Maximum reviews per platform, 1–100 (default: 25)

Returns: Individual review array plus cross-platform summary with sentiment breakdown.

aggregateRatings

Get aggregate ratings summary across all detected platforms — average rating, total review count, and per-platform breakdown.

Parameters:

Parameter	Type	Required	Description
businessUrl	string	Yes	Business URL on a review platform
businessName	string	No	Business name to search for
location	string	No	City/address to disambiguate

Returns: Average rating, total reviews across platforms, platform count, sentiment.

exportReviews

Export scraped reviews in specified format with all review data and metadata.

Parameters:

Parameter	Type	Required	Description
reviews	array	Yes	Array of review objects to export
format	string	No	Export format: json, csv, or markdown (default: json)
filename	string	No	Output filename without extension (default: reviews)

Returns: Export confirmation with format, filename, and review count.

detectFakeReviews

Analyze reviews for potential fake indicators using multiple signal detection methods.

Parameters:

Parameter	Type	Required	Description
reviews	array	Yes	Array of review objects to analyze
threshold	number	No	Minimum fake probability (0–1) to flag as suspicious (default: 0.7)

Detection Signals:

• Missing rating or very short review body
• No review date
• Generic or suspicious reviewer name patterns
• Rating anomalies (5 stars with 10 character review)
• ALL CAPS title
• Excessive punctuation
• URLs in review text
• Repeated character patterns (e.g., "goooood")

Returns: Analysis with probability scores, flags, and list of suspicious reviews.

Input Schema

{
  "operation": "scrapeReviews",
  "businessUrl": "https://www.trustpilot.com/review/airbnb.com",
  "businessName": "Airbnb",
  "location": "Global",
  "platform": "all",
  "maxReviews": 25,
  "exportFormat": "json",
  "fakeReviewThreshold": 0.7
}

Operations

Operation	Description
scrapeReviews	Scrape reviews from specified platform(s)
aggregateRatings	Get aggregate ratings summary
exportReviews	Export reviews to JSON/CSV/Markdown
detectFakeReviews	Analyze reviews for fake indicators

Output Schema

scrapeReviews Response

{
  "operation": "scrapeReviews",
  "input": {
    "businessUrl": "https://www.trustpilot.com/review/airbnb.com",
    "platform": "all",
    "maxReviews": 25
  },
  "summary": {
    "averageRating": 4.3,
    "totalReviewsAcrossPlatforms": 847,
    "platformCount": 3,
    "platforms": [
      {
        "platform": "google",
        "averageRating": 4.5,
        "totalReviews": 420,
        "reviewsScraped": 25
      },
      {
        "platform": "trustpilot",
        "averageRating": 4.1,
        "totalReviews": 327,
        "reviewsScraped": 25
      },
      {
        "platform": "yelp",
        "averageRating": 3.8,
        "totalReviews": 100,
        "reviewsScraped": 10
      }
    ],
    "sentiment": "positive",
    "sentimentBreakdown": {
      "positive": 38,
      "neutral": 8,
      "negative": 4
    },
    "reviewsAnalyzed": 46,
    "timestamp": "2026-05-19T21:30:00.000Z"
  },
  "reviews": [
    {
      "platform": "google",
      "rating": 5,
      "reviewTitle": "Excellent service",
      "reviewBody": "Had a great experience with this business. Fast delivery and friendly staff.",
      "reviewerName": "Alex M.",
      "reviewDate": "2026-02-15",
      "verified": false,
      "helpfulVotes": 3,
      "url": "https://www.google.com/maps/..."
    }
  ],
  "reviewCount": 46,
  "timestamp": "2026-05-19T21:30:00.000Z"
}

detectFakeReviews Response

{
  "operation": "detectFakeReviews",
  "totalReviews": 46,
  "suspiciousCount": 3,
  "cleanCount": 43,
  "threshold": 0.7,
  "fakeProbability": 0.065,
  "suspiciousReviews": [
    {
      "platform": "yelp",
      "rating": 5,
      "reviewerName": "J.",
      "reviewDate": "2026-01-01",
      "fakeProbability": 0.82,
      "flags": ["veryShort", "noDate", "genericReviewer"]
    }
  ],
  "timestamp": "2026-05-19T21:30:00.000Z"
}

Supported Platforms

Platform	URL Pattern	Features
Google	google.com/maps/search/... or place URL	Reviews, ratings, user metadata
Trustpilot	trustpilot.com/review/...	Verified reviews, aggregate rating, helpful votes
Yelp	yelp.com/biz/...	Reviews, ratings, user info
TripAdvisor	tripadvisor.com/...	Reviews, ratings, travel-specific data
Amazon	amazon.com/.../product/... or amzn.com/...	Product reviews, verified purchase badge

Pricing (Pay-Per-Event)

Event	Cost
scrape_reviews	$0.05
aggregate_ratings	$0.03
export_reviews	$0.02
detect_fake_reviews	$0.06
platform_scrape	$0.04 per platform

Example: Scraping reviews across all 5 platforms with 25 reviews each = 5 platform scrapes = $0.20 total.

Error Handling & Retry Logic

This actor implements exponential backoff retry for transient failures:

• Max retries: 3 attempts
• Base delay: 1000ms, doubling each retry (1000ms → 2000ms → 4000ms)
• Max delay: 30 seconds
• Retryable errors: Network timeouts, 502/503/429 responses, connection resets

Failed scrapes return partial results with error annotations per platform rather than failing the entire operation.

Use Cases

• Reputation monitoring — Track ratings and sentiment across platforms over time
• Competitive analysis — Compare review patterns between your business and competitors
• Lead qualification — Gauge customer satisfaction before sales outreach
• Market research — Extract customer pain points and product feedback at scale
• Fake review detection — Identify suspicious reviews for further investigation
• Support ticket triage — Identify negative reviewers for proactive outreach

Quick Start

# Install dependencies
npm install

# Run locally with test input
echo '{"operation":"scrapeReviews","businessUrl":"https://www.trustpilot.com/review/airbnb.com","platform":"all","maxReviews":10}' | INPUT_STORE=1 apify run

# Run with a file
apify run --input-file ./test-input.json

Test Input (test-input.json)

{
  "operation": "scrapeReviews",
  "businessUrl": "https://www.trustpilot.com/review/airbnb.com",
  "platform": "all",
  "maxReviews": 10
}

Architecture

• Standby mode — Spins up on-demand when MCP request arrives, spins down when idle
• MCP protocol — Native MCP server implementation for AI agent integrations
• Readiness probe — Health check at GET / with x-apify-container-server-readiness-probe header
• 90s tool timeout — Prevents long-running operations from hanging
• No external APIs — Uses public HTML scraping, no API keys or credentials required

File Structure

multi-review-scraper-mcp/
├── .actor/
│   └── actor.json          # Actor definition with PPE pricing
├── src/
│   └── main.js             # MCP server implementation
├── Dockerfile              # Container build definition
├── package.json            # Node.js dependencies
└── README.md               # This file

Troubleshooting

Issue	Solution
Empty reviews returned	Business may not have reviews on that platform; try a different URL
Platform not detected	Use direct review page URL (e.g., trustpilot.com/review/example.com)
Rating mismatch	Platform counts may differ due to filtering or rounding
Timeout errors	Reduce maxReviews to 10–25 for large businesses
Fake detection seems off	Adjust threshold parameter (lower = more sensitive)

Build & Deploy

# Syntax check
node --check src/main.js

# Build Docker image
apify build

# Push to Apify platform
apify push

# View on Apify
apify open

License

MIT License — Wilson Davalos