MCP (Model Context Protocol) is an open protocol that standardizes how applications provide context to AI Agents (LLMs). MCP provides a standardized way to connect AI Chatbots and AI Agents to different data sources and tools and allows AI Agents to discover which tools and which data (resources) are available on each MCP Server.

In MCP you have MCP Clients and MCP Servers. The MCP Client uses MCP Servers.

Examples of MCP Clients include Visual Studio, Claude Desktop, OpenAI Playground etc.

Peliqan allows you to build your own custom MCP Server(s) that you can use internally inside your company, or that you can make available to a wider audience for usage. Your MCP Servers built on top of the Peliqan Data Cloud can consume data from the data warehouse, connect directly to business applications etc.

More info on MCP: https://modelcontextprotocol.io/introduction

In the below sections, we will describe how you can set up an MCP server that connects to your Peliqan account, to fetch e.g. data from your data warehouse. We will show how to use your MCP Server in an MCP Client such as OpenAI, Claude Desktop or Visual Studio (VSCode).

Peliqan provides 2 deployment scenarios for your MCP Server:

Remote MCP Server for usage in MCP Clients such as OpenAI playground
Local MCP Server for usage in MCP Clients such as Claude and Visual Studio

In both deployment scenario’s, you implement your own MCP Server methods on the Peliqan Data Cloud, in a custom API endpoint, using a low-code Python script in Peliqan.

Remote MCP Server on Peliqan

In this scenario, your MCP Server runs on the Peliqan Data Cloud, and the MCP Client - e.g. ChatGPT in OpenAI Playground - connects to your Remote MCP Server directly.

Local MCP Server with Peliqan

In this scenario, the MCP Server runs locally on the computer, next to some MCP Client such as Visual Studio or Claude Desktop. The MCP Server is installed by the user on his/her computer. This scenario needs to be used if the MCP Client does not (yet) support Remote MCP.

The local MCP Server on the client communicates with an API endpoint on the Peliqan Data Cloud. You implement your own methods (tools) directly on the Peliqan Data Cloud inside the API endpoint. The locally installed Peliqan MCP Server acts as a proxy to the Peliqan Data Cloud.