Hooks into your Zod schemas to generate valid mocks, boundary violations, and contract test scaffolding without manual payload writing. Exposes tools for creating diverse test data with seeded variance, generating exhaustive union branch violations, and detecting drift between Zod definitions and OpenAPI specs. Also handles schema evolution checks by validating old mocks against new schemas to catch breaking changes before deployment. Outputs contract test boilerplate for Playwright, Jest, Vitest, and MSW. Useful when you need property-based testing data, want to verify API contracts match their runtime validators, or need to regression-test schema migrations against existing fixtures.
An MCP server that turns Zod schemas into mocks, violations, and contract tests — so your AI agent can reason about API contracts without manually crafting payloads.
Zod schemas are runtime code. An AI agent cannot execute them, introspect their constraints, or generate valid/invalid payloads without this layer. The agent also cannot detect when the schema and the OpenAPI docs silently diverged, or whether a schema change breaks existing test fixtures.
| Tool | Arguments | What it returns |
|---|---|---|
generate_valid_mock | schema_code, count? | Valid mock data matching the schema |
generate_mock_variants | schema_code, count?, seed? | N structurally valid but value-diverse mocks — for property-based testing |
| Tool | Arguments | What it returns |
|---|---|---|
generate_boundary_violations | schema_code | Invalid payloads for each constraint: missing fields, type mismatches, min/max, email/uuid/url |
generate_exhaustive_union_violations | schema_code | Per-variant violations for every branch of a z.union() or z.discriminatedUnion() |
| Tool | Arguments | What it returns |
|---|---|---|
introspect_schema | schema_code | JSON Schema representation — for LLM understanding of the contract |
read_schema_from_file | file_path, export_name? | Extracts the Zod schema expression from a TypeScript/JS file |
detect_schema_drift | zod_file_path, schema_export_name, openapi_file_path, openapi_schema_name? | Diffs Zod vs OpenAPI — reports missing_in_openapi, missing_in_zod, type_conflict, required_mismatch |
evaluate_schema_evolution | schema_file_path, schema_export_name, old_schema_content? | Generates mocks from old schema, validates against new — detects breaking changes before tests run |
| Tool | Arguments | What it returns |
|---|---|---|
scaffold_api_contract_test | framework, base_url, endpoint, method, schema_code, test_name? | Contract test boilerplate for Playwright, Jest, Vitest, or MSW |
suggest_contract_fix | schema_code, payload | Validates a JSON payload and explains each violation with a fix suggestion |
npm install -g zod-contract-mock-forge-mcp
.cursor/mcp.json or .vscode/mcp.json){
"mcpServers": {
"zod-forge": {
"command": "zod-contract-mock-forge-mcp"
}
}
}
claude mcp add zod-forge zod-contract-mock-forge-mcp
My schema file is src/schemas/user.ts, exported as UserSchema.
My OpenAPI spec is docs/openapi.yaml.
1. introspect_schema — what are the constraints on this schema?
2. generate_mock_variants — give me 10 diverse valid payloads (seed: 42) for CI reproducibility
3. generate_exhaustive_union_violations — test every branch of the role discriminated union
4. detect_schema_drift — has the Zod schema diverged from the OpenAPI docs?
5. evaluate_schema_evolution — does my schema change break any existing mock data?
generate_mock_variants — 3 diverse valid mocks, seeded for CI:
{
"schema_id": "schema_a1b2c3d4",
"count": 3,
"all_valid": true,
"variants": [
{ "name": "Colleen Rowe", "age": 37 },
{ "name": "Pat Reynolds", "age": 24 },
{ "name": "Veronica Konopelski", "age": 45 }
]
}
detect_schema_drift — field missing in OpenAPI, extra field in Zod:
{
"drift_count": 2,
"drifts": [
{
"field_path": "role",
"drift_type": "missing_in_openapi",
"zod_value": "string",
"openapi_value": null
},
{
"field_path": "email",
"drift_type": "missing_in_zod",
"zod_value": null,
"openapi_value": "string"
}
]
}
evaluate_schema_evolution — new required field breaks existing mocks:
{
"breaking_change": true,
"sample_count": 20,
"invalid_mock_count": 20,
"failure_reasons": [
{
"field_path": "status",
"zod_code": "invalid_type",
"expected": "string",
"received": "undefined",
"affected_mock_count": 20
}
]
}
npm run build # compile TypeScript → dist/
npm run lint # ESLint
npm run format # Prettier --write
npm run format:check # Prettier check (used in CI)
npm test # Vitest
MIT