> ## Documentation Index
> Fetch the complete documentation index at: https://docs.nippy.la/llms.txt
> Use this file to discover all available pages before exploring further.

# MCP — Introduction

> Connect an AI agent to the Nippy engine and operate your platform in natural language

## What is the Nippy MCP

Nippy's MCP (Model Context Protocol) exposes the entire operational engine of the platform as tools that any AI agent can invoke. Instead of opening the console, navigating menus, and exporting reports, you ask directly:

* *"How many active campaigns do we have this week and which one has the best conversion?"*
* *"Create a roulette for the summer campaign with 500 points as the main prize"*
* *"Which products are below minimum stock at the Monterrey center?"*

The agent executes the operations, fetches data, and acts — all in seconds.

## How it relates to the SDK

The SDK and MCP access the same engine. They are two perspectives on the same system:

|                       | SDK                              | MCP                                            |
| --------------------- | -------------------------------- | ---------------------------------------------- |
| **Who uses it**       | The end user (in your app)       | Your team or AI agent                          |
| **How it's accessed** | npm package embedded in your app | MCP client: Claude, Cursor, OpenClaw, Hermes   |
| **What it does**      | The user spins, wins, claims     | The agent creates campaigns, analyzes, manages |
| **Real-time**         | Yes — every spin is synchronous  | Yes — operations happen instantly              |

A bank can use both simultaneously: the SDK runs in the customer's app, the MCP runs in the marketing team's workspace.

## Available verticals

The Nippy MCP exposes six operational areas. Each vertical has its own tools that the agent discovers automatically upon connecting.

<CardGroup cols={2}>
  <Card title="Analytics" icon="chart-bar" href="/en/mcp/analytics">
    Query roulette, supply, course, and attendance data. Your agent builds the query and returns results — no SQL, no Excel.
  </Card>

  <Card title="Supply" icon="boxes-stacking" href="/en/mcp/supply">
    Stock levels by center, withdrawals, low inventory alerts. Create and update from chat.
  </Card>

  <Card title="Roulettes" icon="ticket" href="/en/mcp/roulettes">
    Create and manage gamification campaigns from the agent. Full flow: proposal, inventory validation, publishing.
  </Card>

  <Card title="Flows" icon="route" href="/en/mcp/flows">
    Registration landing pages for campaigns. Configure fields, redirects, and review conversion in real-time.
  </Card>

  <Card title="Learning" icon="graduation-cap" href="/en/mcp/learning">
    Courses, enrollments, exams, and certificates. Create and update without opening the platform.
  </Card>
</CardGroup>

## Compatible clients

The server uses **MCP over HTTP streamable** (JSON-RPC 2.0). It works with any client that implements the protocol:

* **Claude Desktop** — native integration, tools appear in chat
* **Cursor** — agent with direct engine access from the editor
* **OpenClaw** — MCP-compatible local runtime
* **Hermes Agent** (NousResearch) — open source agent with native MCP client

<Card title="Connect in 3 steps" icon="rocket" href="/en/mcp/quickstart" horizontal>
  API key → configure client → first operation. Ready in minutes.
</Card>