Model Context Protocol (MCP) Integration

What is MCP?

The Model Context Protocol (MCP) is an open standard developed by Anthropic that enables AI assistants to securely connect with external data sources and tools. It provides a standardized way for AI models to access systems without custom integrations.

MCP uses a client-server architecture with JSON-RPC 2.0 messaging to create secure, two-way connections between AI applications (like Claude) and your data sources. This eliminates the need for custom API integrations for each AI tool you want to use.

For more details, see the official MCP documentation.

Why MCP in Sidestnd?

Sidestnd already supports HTML and REST interfaces. Even though REST is easy to use for computers, it still requires (a bit of) programming effort. As AI is getting more and more important and companies are starting to build their on-premise AI models, making access to data more convenient is useful. MCP is close to REST, but a little bit easier to use when your goal is to train a model or use inference.

What does MCP offer in Sidestnd?

MCP in Sidestnd focuses on the resources part of the standard. It gives access to the following information:

• discover/list works owned by a specific organization
• work metadata containing URIs for physical files that can subsequently be downloaded
• better understanding of the relationship between an organization and their work

Three Resource Types

1. Organization Details (mcp://organisations/{org-guid})

   • Complete organization information from our Pydantic models
   • Includes all organization metadata and properties

2. Organization Works List (mcp://organisations/{org-guid}/works)

   • List of all works owned by an organization
   • Returns work GUID and name tuples for easy discovery
   • Enables AI to understand what works are available

3. Individual Work Details (mcp://works/{work-guid})

   • Complete work metadata from our Pydantic models
   • Includes URIs for supporting documents and files
   • AI gets context about work type, status, and related resources

How to Use MCP?

MCP Endpoint

All MCP communication happens through:

POST https://<sidestndomain>/mcp/

(or your custom domain if appropriate)

Getting Started

1. Find Your Organization GUID

• Your organization GUID is the key to accessing your works
• This GUID provides both identification and access control
• Keep this GUID secure as it grants access to your organization's data

2. Connect an MCP-Compatible AI Assistant

• Claude Desktop: Configure MCP server connection pointing to our endpoint
• Other MCP Clients: Any tool supporting MCP can connect using the standard protocol

3. Resource Discovery

Once connected, the AI can:

• List available resource types
• Discover works owned by your organization
• Access individual work details and metadata

Access Patterns

Organization-level Discovery:

mcp://organisations/{your-org-guid}/works

Returns list of all works your organization owns.

Individual Work Access:

mcp://works/{specific-work-guid}

Returns complete metadata for a specific work.

Organization Details:

mcp://organisations/{your-org-guid}

Returns your organization's complete profile information.

Security Model

• GUID-based Access: Organization and work GUIDs (32 characters each) provide computational security
• No Traditional Auth: No usernames, passwords, or API keys needed
• Scoped Access: Organizations can only access their own works
• Discovery Protection: Must know organization GUID to discover work GUIDs

Technical Notes

• Protocol: JSON-RPC 2.0 over HTTPS
• Format: All responses are JSON with proper MCP envelope
• File Access: metadata includes references to files (relative URI path, starting with /resources)
• Integration: Reuses existing business logic and Pydantic models
• Performance: Lightweight responses focused on metadata and discovery