CLI tools for web extraction, Amazon product research, and market intelligence via ZooData API
This MCP server defines 25 tools across two functional areas (web extraction and Amazon market analysis). While parameter schemas are present and mostly typed, there are significant quality gaps: (1) Many tool descriptions are adequate but generic, lacking LLM-optimized context on WHEN to use each tool vs alternatives. (2) Output schemas are undocumented, the code shows these are HTTP API wrappers but nowhere visible are the response field definitions that downstream agents need for chaining. (3) Parameter descriptions vary in quality; several lack format/constraint guidance. (4) No error handling patterns visible, tools return raw API responses without recovery guidance. (5) Tool naming is generally good (verb_noun pattern holds), but some tools are near-synonyms (crawl vs map vs interactive) without clear differentiation. (6) Security patterns are absent, API key handling is present but no permission scopes or audit trails visible. Overall, this reads as a thin wrapper over a third-party API with minimal agent-specific polish.
Get Amazon product categories for a keyword
Check API connectivity and authentication status
Get competitor products for a keyword
Crawl a website starting from a URL, following links up to specified depth and limits
Check the status of a running or completed crawl job
Submit a crawl and wait for completion, polling until done or timeout
NO OUTPUT SCHEMAS DOCUMENTED. Not a single tool has a documented response structure. LLMs cannot plan multi-step chains (e.g., categories → market → products) without knowing what fields each returns. This blocks proper tool composition and forces agents to guess.
MISSING PARAMETER CONSTRAINTS & ENUMS. Parameters like 'format', 'scrape_format', 'mode', 'sitemap_mode' lack enum constraints. Free-text parameters invite hallucinated values. LLMs will invent options like 'format=pdf' or 'mode=bestsellers' that fail silently.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | D | 59 | 2026-07-28+ | v2 |
Execute interactive actions on a webpage (clicking, typing, scrolling) and scrape the result
Get detailed metrics for specific keywords on a given date
Get keyword extensions and suggestions based on a query
Get market profile metrics for keywords on a date
Get traffic terms for a product ASIN on a specific date
Get Amazon search results for a keyword on a specific date
Get keyword trend data over a date range
Map a website structure by discovering pages and their relationships
Get Amazon market data for a category
Identify market opportunities for a keyword
Get detailed information about a specific Amazon product by ASIN
Get traffic structure profile for multiple ASINs on a date
Get product traffic trends for specific keywords over a date range
Get product traffic trends for multiple ASINs over a date range
Get product traffic trend profile for an ASIN on a date with window periods
Search for Amazon products by keyword with optional selection mode
Generate a market research report for a keyword
Scrape a URL and extract content in multiple formats (json, markdown, rawHtml)
Search the open web with keyword query and optional filters for domains, time range, and sources
VAGUE OR MISSING DESCRIPTIONS. 16+ tools have descriptions under 60 characters or lack context on WHEN to use them vs similar tools. 'Map a website structure' doesn't explain vs crawl. 'Generate a market research report' doesn't differentiate from 'opportunity'. LLMs will pick wrong tools.
UNDOCUMENTED PARAMETER FORMATS. 'actions' in interactive tool lacks schema for the JSON array structure expected. 'tbs', 'sitemap_mode', 'window_periods', 'include_paths' lack examples or valid value documentation. LLMs will struggle to invoke these correctly.
NEAR-SYNONYM TOOLS WITHOUT DIFFERENTIATION. 'scrape', 'interactive', 'search', 'map', and 'crawl' all extract web data but with unclear boundaries. The descriptions don't guide LLMs on which to call for a given user intent. Similarly, 'report' vs 'opportunity' and 'keyword-detail' vs 'keyword-market-profile' lack distinction.
NO ERROR HANDLING GUIDANCE. No tool documents recovery paths (e.g., 'If crawl times out, try crawl-wait with a longer max_wait'). No tool describes retryability (is this API call idempotent?). Agents cannot self-correct on failures.
WEAK NAMING ON 4+ TOOLS. 'map', 'product', 'report', 'opportunity', 'keyword-extends' lack clear verbs or are too generic. 'product' is ambiguous (create, delete, describe?). 'keyword-extends' uses 'extends' awkwardly (should be 'expand_keywords' or 'get_keyword_suggestions'). 'map' doesn't clarify if it's discovering, navigating, or visualizing.
NO TOOL COMPOSITION GUIDANCE. Tools return responses (presumably) but nowhere is it documented which tools chain together. Do categories → market take category_id from categories output? Does product take ASIN from products output? LLMs must infer field names.
LACK OF PAGINATION/LIMIT GUIDANCE. Many search/crawl tools accept 'limit' but no description states the default, max, or risk of huge result sets. Best practice baselines (50-100 items) are not enforced. An LLM could request limit=10000 and blow the context window.
SECURITY: NO PERMISSION/SCOPE DECLARATIONS. No tool declares what permissions it requires (read:amazon, write:data, etc.). No audit trail or rate-limiting visible. Agents could invoke any tool without privilege checks.