Loading ToolBench
Flights owen merry mcp server
Single tool 'flight-search' has a clear verb-noun name and basic schema with Zod validation. The tool returns unstructured text wrapped in a generic content array, making it difficult for LLMs to extract structured flight data. Error handling provides some context but lacks recovery guidance. No pagination, no result limits documented, and no field-level descriptions for parameters.
Parameter descriptions are too short (10-20 chars vs. 72-char baseline). 'from airport iata code' and 'to airport iata code' lack context on format constraints, examples, or when each is required.
Output schema is undocumented. Tool returns {content: [{type: 'text', text: string}]} but the structure and content of the text field are not specified. LLMs cannot plan downstream operations or extract structured flight data (price, airline, departure time, etc.).
No pagination or result limits documented. Tool may return hundreds of flights with no limit, risking context window exhaustion.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 34 | 2026-07-28+ | v2 |
Error handling lacks recovery guidance. Errors like 'Error fetching flight data: 404' tell the LLM nothing about what to do next.