Schema

Zero-Trust Boundary: Schema Validation

🚀 Value Proposition

Secure AI integrations. We enforce strict Zod 4 validation at every I/O boundary. Our architecture provides deterministic payloads and resilient workflows, avoiding raw exceptions.

• Safe Parsing: Always use Zod 4's .safeParse(input) at I/O boundaries. It avoids throwing exceptions and handles invalid inputs gracefully.

Autonomous Reliability: Intercept and Correct Hallucinations...

We frame Zod validation as a Zero-Trust Data Boundary. We strictly validate every input parameter. We use z.coerce to fix LLM hallucinations, ensuring reliable workflows.

Enforce Reliability: Enterprise Validation Rules

We strictly enforce these schemas at the MCP boundary for secure integrations:

1. Strict Error Responses: Failing tools return structured error responses. Errors extend MySQLMcpError and use formatHandlerError(). Our design prevents unhandled exceptions, empowering autonomous workflows. The schema is: { success: false, error: string, code: string, category: string, suggestion: string, recoverable: boolean }
2. Deterministic Table Queries: Table queries return { exists: false, table: string } for missing tables. This enables graceful fallback handling in AI workflows.
3. Sandboxed File Operations: We bound file interactions using ALLOWED_IO_ROOTS. This prevents directory traversal and ensures isolated, secure execution.
4. Standard Schema Interoperability: Tools use Zod 4's Standard Schema implementation. This supports sophisticated aliases and type coercion while maintaining JSON Schema compatibility.

Build with Confidence: Integration Guide

See the repository source for Zod schemas (src/adapters/mysql/schemas/). We recommend using Zod 4's .safeParse(input) to handle validation errors gracefully without exceptions.

For runtime schema introspection, see the Code Mode documentation. It lets you discover API parameters programmatically using the .schema() introspection method.