Loading ToolBench
Price scraper CLI for Amazon, Google Shopping, Bol.com, and Coolblue. Scrape prices, offers, buy-box, variants, and reviews from your terminal. Stable JSON envelope, MCP server for Claude Desktop / Cursor / Code.
18 tools with complete JSON schemas and descriptions. Naming follows verb_noun pattern (ssc_offers, ssc_search, etc.). Descriptions are concise (50-150 chars) and include credit costs. However, parameter descriptions are minimal, many lack actionable constraints (e.g., 'EAN/GTIN/UPC, 8-14 digits' is present but format validation rules are sparse). Output schemas are not documented in the tool definitions. Error handling is present but generic. Tool annotations (requiresConfirmation) are used for high-cost operations, which is good. No pagination guidance in list-like tools (ssc_history, ssc_offers_results). Missing output schema documentation is a significant gap for LLM planning.
Get the current buy-box winner and price for an EAN. Cost: 1 credit per call.
Show remaining credits and plan info. Cost: 0 credits per call.
Recent API call history for the current key. Cost: 0 credits per call.
Get product information for an EAN. Cost: 1 credit per call.
Match an EAN to a product on Amazon or Bol.com. Cost: 1 credit per call (4 credits with deepsearch). ADVISORY: requiresConfirmation — host SHOULD prompt the user before invocation.
Channel API: acknowledge a match results page by page_token. Cost: 0 credits per call.
Channel API: collect one page of finished match results (max 1000). Cost: 0 credits per call.
Output schemas not documented. Tool definitions lack return type documentation, forcing LLMs to infer response structure and plan downstream calls blindly.
Parameter descriptions lack actionable constraints. 'EAN/GTIN/UPC, 8-14 digits' is present but missing min/max, regex patterns, and format guidance. LLMs cannot validate inputs without explicit constraints.
Pagination guidance missing. ssc_history and ssc_offers_results accept 'limit' but lack documentation of default page size, max results, and next_cursor/offset behavior.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 55 | 2026-07-28+ | v2 |
Channel API: queued/claimed/done/failed counts for submitted match jobs. Cost: 0 credits per call.
Channel API: submit up to 50,000 EANs for Google catalog matching in one country. Costs 1 credit PER EAN. Async: poll match_status, drain match_results, ack pages within ~6h. Cost: 1 credit per EAN. ADVISORY: requiresConfirmation — host SHOULD prompt the user before invocation.
List all seller offers for an EAN on Amazon or Bol.com (legacy sync endpoint). Cost: 1 credit per call.
Channel API: acknowledge an offers results page by page_token. Cost: 0 credits per call.
Channel API: collect one page of finished offer results (max 1000). Cost: 0 credits per call.
Channel API: queued/claimed/done/failed counts for submitted offer jobs. Cost: 0 credits per call.
Channel API: submit up to 50,000 EANs for Google Shopping offers in one country. Costs 1 credit PER EAN. Async: poll offers_status, drain offers_results, ack pages within ~6h. Cost: 1 credit per EAN. ADVISORY: requiresConfirmation — host SHOULD prompt the user before invocation.
Scrape a product page by URL. Cost: 1 credit per call.
Get product reviews for a SKU. Cost: 1 credit per call.
Search Google Shopping by keyword in a country. Cost: 1 credit per call.
Get product variants for a SKU. Cost: 1 credit per call.
Error handling is generic. No recovery guidance in tool descriptions. LLMs cannot self-correct on invalid EAN, invalid country code, or API rate limits without explicit error classification.
Channel API tools (offers_submit, match_submit) lack composition guidance. Descriptions do not explain the async pipeline (submit → status → results → ack) or when to call each step.