features.
Validated against 2,200+ real-world OpenAPI specs and 98,000+ endpoints. This is what you get.
realistic data, not "string"
Responses are generated from your OpenAPI schema with realistic values: names, emails, UUIDs, dates, currencies, addresses. No handwritten fixtures. No "lorem ipsum" payloads that lie to your client.
multiple APIs on one port
Drop multiple OpenAPI specs into one folder and Mockzilla mounts each as its own service. Stripe, Twilio, SendGrid, and your own API all reachable on a single port, each under its own URL prefix.
upstream proxy with mock fallback
Point a service at a real backend and Mockzilla forwards requests. If the upstream endpoint doesn't exist yet, it falls back to a generated mock. Mix real and mocked endpoints in the same server while your backend catches up.
latency and error injection
Inject delays and synthetic errors with one flag. --latency 200ms, --errors p5=500,p10=503. Test how your client actually behaves when the network gets weird.
request validation
Every incoming request is validated against the OpenAPI spec. Wrong content type, missing required field, malformed payload? You get a real 4xx, not a silent pass. Catch contract drift before production does.
selective overrides
Auto-generate hundreds of endpoints, hand-craft the few that need exact responses. Drop a JSON file at users/{id}/index.json and that endpoint is now static. Everything else stays auto-generated.
custom middleware
Need behavior the spec can't express? Rewrite requests, mutate responses, inject auth, persist state. In codegen mode you get typed Go handlers wired into the same engine.
.mockz portable packages
Run mockzilla pack against a service folder and you get a single .mockz archive holding specs, static responses, context, and config. Ship it. Host it. Embed it. mockzilla file.mockz and you're running.
MCP server for AI tools
The Mockzilla MCP server lets Claude Code, Cursor, and Gemini CLI spin up mock APIs directly from any OpenAPI definition. Point the agent at a spec and iterate on client code without leaving the editor.
portable or codegen
Two modes, same engine. Portable: point at specs, run, done. Codegen: generate typed Go handlers with custom logic for advanced use cases. Templates for both at github.com/mockzilla.
cloud, when you want it
Mockzilla Cloud runs the same engine as a hosted service. Push to GitHub, get a simulation URL per branch, share with your team. Per-PR mocks, zero-config GitHub Actions, multi-region. Same .mockz files.
open source
MIT licensed. Validated against 2,200+ real-world OpenAPI specs. github.com/mockzilla/mockzilla.