AI Integration (MCP)

Mockzilla is built for the era of AI. It supports the Model Context Protocol (MCP), which means you can give any modern AI agent direct access to your Mockzilla server tools.

🤖 AI-First: Agent Skills

While raw MCP tools are powerful, you can supercharge your AI agent with specialized Mockzilla Skills. These are expert instruction sets that teach your agent exactly how to follow our best practices for spec-driven development.

Universal Installation

Install our expert skills into any compatible agent using the universal npx skills tool.

npx skills add github.com/andrecrjr/mockzilla

Available Experts:

Mock Maker: High-fidelity mocks with JSON Schema + Faker.
Workflow Architect: Stateful scenarios and business logic.
Spec Translator: Bootstrapping from OpenAPI/Specs.
Logic Doctor: Forensic debugging and state repair.

Orchestrating with AI Agents

Mockzilla provides the critical "Ground Truth" infrastructure for modern AI orchestrators like Claude Opus, GPT, and Gemini.

🏗️ The Orchestrator

High-reasoning AI assistants manage complex development lifecycles. They use Mockzilla to:

Spec-to-Mock: Convert project requirements into functional API simulations.
Traffic Audit: Compare real-world traffic captured by the Mockzilla Extension with existing specifications.
Team Sync: Push and pull mocks between local environments and shared servers.

🦎 The Edge Agent

Minimalist agents designed for edge devices can leverage Mockzilla for:

API Heartbeats: Constantly monitor mock consistency during development.
Remote Relay: Act as a lightweight proxy for Mockzilla tools in resource-constrained environments.

Why use Mockzilla with AI?

Instead of clicking through the dashboard yourself, you can just ask your assistant:

"Create a new mock for /api/products."
"Generate a realistic JSON schema for an e-commerce order."
"Test the /users endpoint and tell me if it matches the schema."
"Build a checkout workflow that updates the inventory table."

Quick Connection

To connect Mockzilla to your preferred AI assistant, use the following configuration details:

Mockzilla exposes MCP over the Streamable HTTP transport at /api/mcp, using stateless JSON responses for broad compatibility with current MCP clients.

MCP Server URL

For web-based or editor-integrated agents, add this URL to your MCP settings:

http://localhost:36666/api/mcp

Local Configuration

For desktop-based agents that require a local stdio server entry, add Mockzilla through mcp-remote. The local endpoint uses http://, so include --allow-http. Do not force sse-only: Mockzilla uses the newer single-endpoint Streamable HTTP transport, not the deprecated two-endpoint HTTP+SSE transport.

Local HTTP example:

{
  "mcpServers": {
    "mockzilla": {
      "command": "npx",
      "args": [
        "-y",
        "mcp-remote",
        "http://localhost:36666/api/mcp",
        "--allow-http"
      ]
    }
  }
}

HTTPS example, without --allow-http:

{
  "mcpServers": {
    "mockzilla": {
      "command": "npx",
      "args": [
        "-y",
        "mcp-remote",
        "https://mockzilla.example.com/api/mcp"
      ]
    }
  }
}

Use localhost for local development. A LAN IP such as 192.168.x.x only works when the MCP client runs on another machine and the Mockzilla host permits inbound traffic to port 36666.

Pro Tip: Ask your AI assistant to design the perfect API response while you focus on building the UI. It's like having a backend engineer built into your editor!