A step-by-step demo tutorial showcasing how to use the new MCP (Message Control Protocol) feature to add powerful tool extensions to your AI agents.
Input schemas exist but parameters lack descriptions (city and message have null descriptions). No output schema is documented. Error handling returns plain text rather than structured guidance. Tool names are acceptable (verb_noun pattern) but insufficient to compensate for missing documentation.
Parameter descriptions are null. Both tools define input schemas with z.string() but provide no descriptions for 'city' and 'message' parameters. LLMs cannot infer parameter meaning from names alone.
No output schema documented. Tools return {content: [{type: 'text', text: ...}]} but this structure is not declared in tool definitions. LLMs cannot plan downstream calls without knowing return fields.
Error responses are unstructured text. Both tools catch errors and return plain text messages like 'Failed to fetch temperature for {city}: {err.message}'. LLMs need actionable recovery guidance, not raw error strings.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 31 | 2026-07-28+ | v2 |
No input validation or constraint documentation. getCityTemperature accepts any string for 'city'; testGroq accepts optional message with no length limits. No enums, patterns, or min/max constraints defined.