A Model Context Protocol (MCP) server that provides access to FetchSERP API for SEO analysis, SERP data, web scraping, and keyword research. Supports both stdio and HTTP transport modes.
Server provides 13 well-named SEO/web-scraping tools with reasonable coverage. All tools have descriptions (10-50 words, averaging 12 chars, quite short). All input schemas are present with type definitions and parameter constraints. However, descriptions lack depth for LLM decision-making: most are single-sentence summaries under 50 characters with no guidance on when to use the tool, what it returns, or dependency hints. Output schemas are not documented, LLMs cannot anticipate response structure. No tool annotations (readOnlyHint, destructiveHint, idempotentHint) despite clear read-only operations. Error handling exists but lacks recovery guidance and categorization. Parameter relationships are undocumented (e.g., no indication that get_keywords_suggestions requires either url OR keywords). Mutually exclusive parameters lack explicit constraints.
Check if a domain is indexed for a given keyword
Get backlinks for a given domain
Retrieve emails from a given domain
Get domain info including DNS records, WHOIS data, SSL certificates, and technology stack
Get domain ranking for a given keyword
Get search volume for given keywords
Descriptions are extremely brief (12 chars average) and lack LLM-decision context. 'Get backlinks for a given domain' tells LLM WHAT but not WHEN to use it vs similar tools, what fields to expect, or prerequisites. Baseline for A-grade tools is 50-200 chars with rationale.
Output schemas completely undocumented. LLMs cannot plan multi-step workflows or extract required downstream IDs (e.g., does get_backlinks return domain_id, url_id, link_count?). Critical for tool chaining.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | C | 61 | 2026-07-28+ | v2 |
| 2026-03-09 | F | 0 | - | v1 |
Get keyword suggestions based on a url or a list of keywords
Generate long-tail keywords for a given keyword
Get Moz domain analysis data
Scrape a domain
Scrape a web page without JS
Scrape a web page with custom JS
Scrape a web page with custom JS and proxy
No tool annotations (readOnlyHint, destructiveHint, idempotentHint) despite clear read-only semantics. All tools are read-only, add readOnlyHint: true to signal safety and enable smarter agent planning.
get_keywords_suggestions has undocumented mutual exclusivity: url and keywords are both optional, but unclear which combination is valid (url alone? keywords alone? both?). Parameter relationship documentation missing.
No error recovery guidance in descriptions. If a domain is not found, scrape fails, or API quota is exceeded, LLM has no direction on retry strategy, fallback tools, or user remediation. All error responses must guide next action.
scrape_webpage_js and scrape_webpage_js_proxy have identical names and descriptions but different implementations (one with proxy support). LLM will struggle to distinguish which to use. The difference should be explicit in naming or routing.
No guidance on limits or pagination. scrape_domain accepts max_pages up to 200, does pagination exist? How many results does get_backlinks return? Large unbounded responses risk context window explosion.