MCP servers for AWS Connect infrastructure and knowledge: CDK docs reference and Connect research tools (docs, blogs, re:Post)
Five tools with clear action verbs (search_*, get_*) and reasonable descriptions (100-250 chars). All have documented input schemas with types and descriptions. However, output schemas are not visible in the provided source code, only input parameters are documented. Parameter descriptions are adequate but lack constraint details (ranges, formats, enums). No evidence of error handling guidance, recovery paths, or actionable error messages. Tool composition is sound (each does one thing), but descriptions could be more LLM-optimized per the 50-200 char baseline.
Fetch and parse a single CDK construct reference page. Use this when authoring or reviewing a CDK construct and you need the canonical property list (name, type, required, description) and the source doc URL — beyond what the ``search_cdk_docs`` one-liner snippets provide.
Search AWS blogs and return formatted text results.
Search AWS documentation and return formatted text results.
Search AWS re:Post questions / articles / knowledge-center.
Search the AWS CDK (Python) API reference. Best for finding the right construct or property when authoring CDK Python infrastructure for Connect work. Scoped to the CDK for Python reference (``docs.aws.amazon.com/cdk/api/v2/python``), which covers the four in-scope construct libraries: ``aws_cdk.aws_connect``, ``aws_cdk.aws_lex``, ``aws_cdk.aws_wisdom``, and ``aws_cdk.aws_bedrockagentcore``. Pair with ``get_cdk_construct_doc`` when you need the full property list for a specific construct.
Output schemas not documented. Tool descriptions state what is returned (e.g., 'formatted text results', 'canonical property list') but the actual response structure (fields, types, pagination) is not visible in source code. LLMs cannot plan downstream tool calls or extract data without knowing response shape.
search_aws_docs has minimal description ('Search AWS documentation and return formatted text results.'). At 56 characters, it lacks context for when to use this vs search_aws_blogs or search_aws_repost. No guidance on scope, result quality, or prerequisites.
No error handling guidance visible. Descriptions do not explain what happens on failure (e.g., 'If no results found, try a broader query' or 'Invalid library name, must be one of: aws_cdk.aws_connect, aws_cdk.aws_lex, aws_cdk.aws_wisdom, aws_cdk.aws_bedrockagentcore'). LLMs cannot self-correct without recovery hints.
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 48 | <=2025-11-25 | v2 |
Parameter constraints not fully specified. 'limit' and 'page_size' parameters lack min/max bounds. 'blogs' parameter (array) has no guidance on valid blog names or format. 'tag_ids' (array) has no documentation of valid tag IDs. LLMs will guess or pass invalid values.
get_cdk_construct_doc 'library' parameter default is null with fallback behavior ('each in-scope library is tried until one resolves'). This is reasonable, but the description does not explicitly list the four valid library names inline, LLMs must infer from the description text. Enum constraint would be clearer.