Exposes scrapling-fetch-mcp tools via Streamable HTTP transport for fetching and processing web pages.
Two web-fetching tools with basic descriptions and parameter documentation. Both tools are read-only, low-risk operations. However, critical gaps exist: (1) descriptions are below optimal length for LLM reasoning (s_fetch_page: 84 chars, fetch: 51 chars, below the 194-char baseline); (2) parameter descriptions lack detail on format constraints and when to use each tool; (3) no output schema is documented, forcing LLMs to guess the structure of returned content; (4) no error handling guidance; (5) naming is somewhat redundant (s_fetch_page vs fetch creates confusion rather than clarity). The tools are functional but lack production-grade refinement.
Fetch a webpage and return its content.
Fetch a webpage and return its content as markdown, HTML, or text.
Ambiguous tool naming: 's_fetch_page' vs 'fetch' creates confusion. No clear distinction in tool purpose or when to use one over the other. Both perform the same operation with slightly different parameter sets.
Output schema is completely undocumented. Tool descriptions state 'returns content in requested format' but do not specify: is it a string? object with metadata? Does markdown mode include any structured elements? LLMs cannot plan downstream operations without knowing output structure.
Tool descriptions are too brief (51-84 chars vs 194-char baseline). s_fetch_page: 'Fetch a webpage and return its content as markdown, HTML, or text.' lacks context on when to select this tool, what markdown conversion does, or error handling. fetch description is even shorter and lacks differentiation from s_fetch_page.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | F | 45 | 2026-07-28+ | v2 |
| 2026-03-09 | F | 39 | - | v1 |
Parameter descriptions lack actionable constraints. 'browser' parameter says 'currently only camoufox is supported' but does not explain: why is this parameter exposed if only one option exists? What happens if an LLM passes 'chromium' or 'firefox'? Is this parameter future-proofing or user-selectable?
No error handling guidance. Tool docstrings do not document: what happens on network timeout? Invalid URL? SSL certificate errors? LLM receives raw error strings ('Error fetching {url}: {str(e)}') with no recovery guidance.
fetch tool signature lacks documentation explaining why it omits the 'browser' parameter. LLM reasoning cannot determine: should I use fetch or s_fetch_page? When is browser selection relevant? This creates tool selection ambiguity.