shadcn-mcp has 5 tools with basic schemas and descriptions, but significant gaps limit production readiness. All tools have descriptions (10-60 chars, mostly short) and input schemas with type definitions. However, parameter descriptions are sparse or missing entirely. list_components and list_blocks lack parameter descriptions; get_component and get_block have minimal guidance on expected formats. No output schemas documented. Error handling is minimal, tools return plain text on failure with no recovery guidance. No tool annotations (readOnlyHint, etc.). The server reads from GitHub and caches results, which is safe, but lacks pagination, result limits, and structured output. Average tool score: 52/100.
Get source code for a shadcn/ui v4 block.
Get source code for a shadcn/ui v4 component.
List all shadcn/ui v4 blocks (pre-built page sections).
List all shadcn/ui v4 components.
Search components and blocks by name or keyword.
No output schemas documented. Tools return TextContent with unstructured text. LLMs cannot parse results reliably or chain downstream calls. E.g., list_components returns a formatted string, not a structured array of {name, type} objects.
Parameter descriptions missing or incomplete. 'category' in list_blocks has a description, but 'query' in search lacks guidance on what to search (name? keyword? both?). LLMs cannot infer intent from bare parameter names.
No pagination or result limits. list_components and list_blocks return all results as plain text. If the registry grows to hundreds of components, responses will bloat context and degrade LLM reasoning. No limit parameter, no cursor, no total count.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 55 | 2026-07-28+ | v2 |
Error handling returns plain text with no recovery guidance. E.g., 'Component not found' gives no hint to call search(). Errors should suggest next steps: 'Component "xyz" not found. Try search(query="xyz") to find similar components.'
No tool annotations. Tools lack readOnlyHint, destructiveHint, or idempotentHint. All tools are read-only and safe to retry, but LLMs cannot infer this from the schema alone. Annotations would improve agent planning.