jcorpac MCP Tool Design
A tool is only as good as its description. AI models depend on clear JSON Schema to function.
Tool Definition
- Name: Concise, snake_case (e.g.,
calculate_risk_score). - Description: Detailed explanation of what the tool does and when to use it.
- Input Schema: Strict types and required fields using JSON Schema.
Execution Logic
- Safety: Validate inputs before processing.
- Feedback: Provide verbose success messages or actionable error reports.
- Idempotency: Ensure running the tool multiple times with the same input has predictable results.
Best Practices
- Schema Evolution: Maintain backward compatibility.
- Context-Awareness: Use tools to supplement the model's knowledge, not replace its reasoning.