Loading ToolBench
MCP server exposing 12 web scrapers as tools for AI agents (ChatGPT, Claude, Cursor). Reddit, eBay, Amazon, Google Maps, Yelp, YouTube, TikTok, Indeed, Trustpilot, contact finder, SaaS pricing, gmaps reviews.
12 tools with complete input schemas and descriptions. Naming is action-oriented (reddit_scrape, amazon_search, etc.). Descriptions are detailed (150-300 chars) with use cases and examples, exceeding the 10-1024 char baseline. All parameters have type definitions and descriptions. However, output schemas are entirely undocumented, LLMs cannot predict response structure for downstream chaining. No error handling guidance. No tool annotations (readOnlyHint, etc.). Parameter descriptions include example values (e.g., 'e.g. ["technology", "SaaS"]'), which LLMs may reuse literally. Missing idempotency and retry guidance for rate-limited services (Reddit RSS fallback, eBay/Amazon blocks).
Search Amazon for products or look up specific ASINs/product URLs. Returns title, price, rating, review count, ASIN, image URL and Prime status. Use for price comparison, product research, and competitor monitoring. Supports US/UK/DE/FR/IT/ES/CA/AU/JP/IN marketplaces. Example input: {"searchQueries": ["wireless earbuds"], "marketplace": "amazon.com", "maxResultsPerQuery": 20}
Find business contact information (email, phone, LinkedIn) for companies and people. Returns name, title, email, phone, LinkedIn URL and company. Use for lead generation, sales prospecting, and recruitment. Example input: {"companies": ["Acme Corp"], "maxResultsPerCompany": 20}
Search eBay listings, including completed/sold listings for real market-price history. Returns title, price, condition, seller, sold count and auction status. Use soldOnly:true to research what items actually sell for (resale/arbitrage analysis). Example input: {"searchQueries": ["vintage seiko watch"], "soldOnly": true, "maxResultsPerQuery": 30}
Extract customer reviews for specific Google Maps places. Returns reviewer name, star rating, full review text, date and owner responses. Use after google_maps_search to dig into reputation/voice-of-customer for a business. Example input: {"placeUrls": ["https://www.google.com/maps/place/..."], "maxReviewsPerPlace": 30, "sortBy": "newest"}
Output schemas completely undocumented. LLMs cannot predict response structure (fields, types, pagination) for downstream tool chaining or data extraction. Forces agents to guess or fail on unexpected formats.
No error handling guidance. Tools that scrape external sites (Reddit, eBay, Amazon, TikTok) are rate-limited and may fail, but descriptions do not explain retry logic, fallback behavior (e.g., Reddit RSS), or recovery steps for LLMs.
Parameter descriptions include example values (e.g., 'e.g. ["technology", "SaaS"]', 'e.g. ["standing desk", "ergonomic chair"]'). LLMs frequently reuse example values literally in real calls, causing failures or unintended behavior.
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 58 | <=2025-11-25 | v2 |
Find local businesses on Google Maps by category + location. Returns name, address, phone, website, rating, review count and GPS coordinates. The go-to tool for building local lead lists or analyzing local competition. Example input: {"searchQueries": ["dentists in Miami FL"], "maxResultsPerQuery": 25}
Search Indeed job listings. Returns job title, company, location, salary (if posted), job description snippet, and apply URL. Use for job market research, salary benchmarking, and hiring trend analysis. Example input: {"searchQueries": ["senior software engineer in San Francisco, CA"], "maxResultsPerQuery": 25}
Get Reddit posts from subreddits, keyword searches across all of Reddit, or comments from specific post URLs. Returns title, score, author, body text, comment count and (optionally) nested comments — ideal for sentiment analysis, market research, and finding what people complain about. Note: when Reddit rate-limits, the scraper transparently degrades to Reddit RSS feeds, so very new comments may be partial. Example input: {"subreddits": ["smallbusiness"], "maxPostsPerSubreddit": 25, "sortBy": "top", "timeFilter": "week"}
Scrape SaaS pricing pages to extract pricing tiers, features, and billing models. Returns plan name, price, billing cycle, features list and free trial info. Use for competitive pricing analysis and feature benchmarking. Example input: {"urls": ["https://www.example.com/pricing"], "extractFeatures": true}
Scrape TikTok creator profiles, their videos, or discover trending videos by hashtag. Returns followers, total likes, and per-video views, likes, comments, shares, hashtags and music. Use for influencer discovery and trend analysis. Example input: {"profileUrls": ["@khaby.lame"], "maxPostsPerProfile": 10}
Extract customer reviews from Trustpilot for any company. Returns reviewer name, star rating, review title, full text, date and company response. Use for reputation monitoring and voice-of-customer analysis. Example input: {"companyUrls": ["https://www.trustpilot.com/review/example.com"], "maxReviewsPerCompany": 30}
Search Yelp businesses by category + location through the official Yelp Fusion API (reliable, no scraping blocks). Returns rating, address, phone, categories, price level, GPS and supported transactions. Requires a free Yelp Fusion API key (5000 calls/day at https://docs.developer.yelp.com/) passed as yelpApiKey. Example input: {"searches": ["ramen in San Francisco, CA"], "yelpApiKey": "YOUR_KEY", "maxResultsPerSearch": 20}
Get YouTube channel stats and recent videos, or metadata for specific video URLs. Returns subscribers, total views, and per-video title, view count, likes and publish date. Use for influencer vetting and content research. Example input: {"channelUrls": ["https://www.youtube.com/@mkbhd"], "maxVideosPerChannel": 15}
No tool annotations (readOnlyHint, destructiveHint, idempotentHint). All tools are read-only scrapers, but this is not declared in the schema. Agents cannot infer safety properties or retry behavior.
No pagination guidance in descriptions. Tools returning lists (reddit_scrape with includeComments, google_maps_search, etc.) do not document how to handle large result sets or whether pagination is supported.