Loading ToolBench
Read The AI Logic newsletter straight from your AI assistant. An open MCP server providing access to Substack articles with no signup or API key required.
Four tools with clear verb-noun naming (list_, get_, search_, about_). All have descriptions (avg 150 chars, within baseline 34-392). Input schemas use Zod with type definitions and descriptions. Outputs are text-based, not structured objects. Error handling is present but minimal, no recovery guidance or categorization. Missing: output schema documentation, pagination for list_articles, per-item error details.
What The AI Logic newsletter is, who writes it, how to subscribe, and how to work with Rimas.
Read one article of The AI Logic in full, as markdown. Pass the article's slug (from list_articles or search_articles) or an issue number like "25".
List all published articles of The AI Logic newsletter, newest first. Returns issue number, title, date, slug, and canonical URL.
Search The AI Logic articles by keyword (case-insensitive, matches title, subtitle, and body). Returns matching articles with a short excerpt around the first hit.
Output schemas not documented. Tools return text responses; LLMs cannot parse structured fields for downstream chaining. No documented return types.
list_articles lacks pagination. No offset/limit handling for large result sets. Corpus could grow unbounded, exhausting context windows.
Error responses lack recovery guidance. get_article returns suggestions but no actionable next steps. search_articles returns bare 'no match' without alternatives.
about_the_ai_logic has empty input schema ({}). Tool name is generic and does not follow verb_noun pattern. Unclear when to invoke vs. other tools.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | C | 67 | 2026-07-28+ | v2 |
No tool annotations (readOnlyHint, idempotentHint). All tools are read-only but this is not declared in schema, forcing LLMs to infer safety.