MCP server that scrapes Strands Agents documentation and indexes it in Elasticsearch for use with Amazon Q
This server exposes 5 tools for searching and retrieving Strands Agents documentation. However, significant quality gaps are evident. Tool definitions lack consistency (search_documentation appears twice with minor description variations), parameter descriptions are minimal, output schemas are entirely undocumented, and error handling guidance is absent. The tools are read-only and low-risk, but the definitions would not guide an LLM effectively through reasoning or recovery. Only 1 of 5 tools has adequate naming clarity; most lack descriptions that explain WHEN to use them vs. similar tools.
Get a specific document by its URL
Get overview of available documentation sections with aggregation data.
List all available documentation sections
Search Strands Agents documentation.
Search through Strands Agents documentation
Duplicate tool registration: 'search_documentation' appears twice (tools #1 and #3) with nearly identical schemas but slightly different descriptions. This creates ambiguity for LLMs and likely represents a copy-paste error during registration.
Output schemas are not documented anywhere in the provided source. LLMs cannot predict what fields search_documentation, get_documentation_sections, get_document_by_url, or list_sections will return. This forces the LLM to guess at response structure and plan downstream calls blindly.
Tool descriptions are generic and lack WHEN/WHY guidance. 'Search Strands Agents documentation' vs. 'Search through Strands Agents documentation' do not explain when to call search_documentation instead of get_document_by_url or list_sections. Descriptions average ~45 chars; rubric baseline is 194 chars for production tools.
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | F | 46 | <=2025-11-25 | v2 |
| 2026-03-09 | F | 41 | - | v1 |
Parameter descriptions are minimal or missing. 'max_results' has a description ('Maximum number of results to return (default: 10)'), but 'section_filter' is marked '(optional)' with no guidance on valid values or formatting. LLMs cannot infer which sections exist or how to filter them.
No error handling or recovery guidance documented. If a search returns no results, if a URL is invalid, or if section_filter is malformed, how should the LLM recover? No error examples or actionable messages are provided.
get_document_by_url accepts a 'url' parameter with no validation guidance. What constitutes a valid URL? Should it be absolute or relative? Does it accept http:// or only https://? No constraints stated.
No pagination parameters documented for search_documentation. The tool accepts 'max_results' (default 10), but does not mention how to retrieve results beyond the first page. Does it support 'offset', 'page', or 'cursor'? If not, agents searching for less-common topics may miss results.