Zod & StandardSchema

Core is 6.26KB gz without Zod; Zod is opt-in to keep lean.

Enable

import 'simple-webmcp/zod'; // side-effect registers Zod → JSON converter
import { z } from 'zod';
import { webmcp } from 'simple-webmcp';

webmcp(fn, { schema: z.object({ query: z.string().min(1) }) });
webmcp(fn, { fields: { query: z.string().describe('Name') } });

Without the side-effect import, StandardSchema schema falls back to inferred/runtime placeholder and per-field Zod falls back to {type:'string'} — tests still pass but less accurate.

Helpers

import { zodToJsonSchema, convertZodDef } from 'simple-webmcp/zod';
zodToJsonSchema(z.string().describe('x')); // → {type:'string', description:'x'}

Supports ZodString (checks min/max/regex/email/url/uuid), ZodNumber (min/max/int), ZodEnum, ZodArray, ZodObject, ZodOptional/Default/Nullable/Effects, unions etc. Valibot/ArkType via StandardSchema validate will pass through when converter not matched (fallback).

Whole vs per-field

// whole
webmcp(fn, { schema: z.object({ a: z.string(), b: z.number().optional() }) });
// per-field
webmcp(fn, { fields: { a: z.string(), b: z.number() } });

Whole schema wins; fields patches descriptions/min/max.