Both tools have minimal definitions. Tool names follow verb_noun pattern (get_*), which is correct. However, descriptions are extremely brief (under 50 chars each) and lack context about when to use each tool or what they return. No input schemas are visible, both tools register with empty input objects `{}`. No output schemas are documented. No parameter descriptions exist (though both tools have zero parameters). Error handling is minimal: tools return generic 'No sarcastic quotes found' messages without recovery guidance. The server fetches from an external API (sarcasmapi.onrender.com) but provides no timeout, retry, or failure context to the LLM.

• Add explicit input schemas (even if empty) with proper JSON Schema format and type declarations. Document any optional parameters or constraints.
• Expand tool descriptions to 100 - 200 characters. Explain: (1) What the tool does, (2) When to use it (all quotes vs random), (3) What structure it returns (array of SarcasticQuote objects with id, sarcasm, createdAt, updatedAt fields).
• Return structured output instead of text blocks. Replace `{ type: 'text', text: JSON.stringify(...) }` with proper MCP content blocks that expose the SarcasticQuote array schema.
• Add timeout (e.g., 5 seconds) to axios calls. Catch network errors and return actionable messages: 'API timeout. The sarcasm service may be down. Retry in a few seconds.' instead of silent failures.
• Implement per-item error reporting for get-all-sarcastic-quotes. If the API returns partial data, return what succeeded and note what failed, rather than failing entirely.