schema-validate

Validate JSON or YAML against a JSON Schema and get back the exact field path of every failure — instead of shelling out to ajv-cli / python -m jsonschema and parsing their text. Built on ajv (draft-07 & 2020-12).

validate({ schema, data: { users: [{ email: 5 }] } })
→ { valid: false, error_count: 1,
    errors: [ { path: "/users/0/email", message: "must be string", keyword: "type" } ] }

Tools

• validate(schema, data, data_format?) — schema/data as objects or strings. Returns { valid, error_count, errors: [{ path, message, keyword, allowed?, missing? }] }. Set data_format: "yaml" when data is a YAML string.
• validate_file(schema_path, data_path) — validate one file against another (format auto-detected by extension).
• check_schema(schema) — is the schema itself valid (compiles against the meta-schema)?

Error paths use JSON Pointer (/users/0/email). allErrors is on, so you get every failure at once. format keywords (email, date-time, uri, …) are enabled.

Setup

npm install
npm test

claude mcp add schema-validate -- node /abs/path/schema-validate/dist/server.js
# CLI: agent-validate schema.json config.yaml

Honest limits

• JSON Schema only (draft-07 by default; 2019-09/2020-12 auto-selected from the schema's $schema). Not other schema languages (Joi, Zod, protobuf).
• strict: false — unknown keywords are ignored rather than erroring, so odd schemas still run.
• YAML is parsed to plain JS then validated; YAML-specific tags aren't schema-checked.