CrestApp Mcp client and server

🔒 You must be logged in as an Administrator or Editor to listen to this audio.

Based on the provided CrestApps Orchard Core repository files and documentation, here is a detailed breakdown of what MCP is, what it is for, and how to use it within the project.

What is MCP?

MCP stands for Model Context Protocol. It is an open standard (originally introduced by Anthropic) that provides a universal, secure way for AI models and assistants to connect to external data sources, tools, and prompts.

In the CrestApps AI ecosystem for Orchard Core, the MCP integration allows Orchard Core to seamlessly share its own capabilities with external AI tools, or augment its own AI agents with external capabilities.

What is it for?

The MCP integration in CrestApps serves a dual purpose, acting as both a client and a server:

  1. Orchard Core as an MCP Server (CrestApps.OrchardCore.AI.Mcp): It exposes Orchard Core’s internal AI tools, predefined prompts, and data resources to external MCP-compatible clients (such as Claude Desktop, Cursor, or other custom AI agents). For example, an external AI could read Orchard Core content items, access FTP/SFTP files via the CrestApps.OrchardCore.AI.Mcp.Resources.Ftp and Sftp modules, or trigger Orchard Core workflows.
  2. Orchard Core as an MCP Client: It allows Orchard Core's own internal AI profiles and chat interactions to connect to external MCP servers. This instantly expands your Orchard Core AI agents' capabilities by granting them access to third-party tools, remote databases, or external file systems without needing to write custom integration code.

How to use it?

1. Using Orchard Core as an MCP Client

To allow your Orchard Core AI agents to consume external tools:

  • Create a Connection: Navigate to the MCP Connections section in the Orchard Core admin dashboard (managed by Controllers/ConnectionsController.cs).
  • Choose a Transport: Configure the connection using either SSE (Server-Sent Events) for remote HTTP-based MCP servers or STDIO (Standard Input/Output) for local command-line based MCP servers.
  • Attach to a Profile: Edit an AI Profile or Chat Interaction and attach the MCP connection. The AI agent will now dynamically discover and use the tools provided by that external server.

2. Using Orchard Core as an MCP Server

To expose Orchard Core data and tools to the outside world:

  • Enable the Feature: Turn on the MCP Server feature within your Orchard Core modules.
  • Expose Prompts and Resources: Use the admin UI to define McpPrompt and McpResource records. You can expose specific content items, static files, or even remote FTP/SFTP directories.
  • Connect your External Client: Point your external AI client (like Claude) to your Orchard Core's well-known MCP SSE endpoint. You can secure this endpoint using API Key authentication or OAuth2.