MCP server for Dash, the macOS documentation browser
Three tools with clear verb-noun naming (list_, search_, get_). All have descriptions (194 - 250 chars, within baseline 34 - 392). Input schemas present with Zod validation and type constraints. However, output schemas are not explicitly documented in the code, responses are text-wrapped without structured type declarations. Error handling includes recovery guidance ('Run list_installed_docsets to see available docsets') but lacks categorization (retryable vs fatal). Token truncation logic is present but not surfaced in parameter constraints. Missing tool annotations (readOnlyHint, idempotentHint) despite all tools being read-only.
Load and convert a documentation page to Markdown. Extracts the relevant section if an anchor is provided, otherwise returns the full page with navigation stripped.
List all installed documentation sets in Dash. An empty list is returned if the user has no docsets installed. Results are automatically truncated if they would exceed 25,000 tokens.
Search for documentation across docset identifiers and snippets. Results are automatically truncated if they would exceed 25,000 tokens.
Output schemas not documented. Tools return text-wrapped JSON without explicit type definitions. LLMs cannot infer response structure for downstream chaining.
Missing tool annotations. All three tools are read-only and idempotent, but lack readOnlyHint and idempotentHint annotations in schema.
Error responses lack categorization. Errors are returned as text but do not indicate whether they are retryable, user-fixable, or fatal.
get_documentation_page parameter 'load_url' lacks format/pattern constraints. No description of expected URL format or source (must come from search results).
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | C | 68 | <=2025-11-25 | v2 |
Token truncation logic is implicit. Descriptions mention '25,000 token limit' but no parameter controls truncation behavior or allows users to request full results.