- 7 minutes to read

What is the Nodinite MCP Endpoint?

The Nodinite MCP Endpoint is the Core Service that exposes your Nodinite environment through the open Model Context Protocol (MCP). AI clients use it to read the Repository Model, Monitor Views, Log Views, and Articles of your environment with typed tools instead of database access. In the Nodinite Portal and the generated installation package, the same component is called the MCP Server (Nodinite.Core.McpServer).

The MCP Endpoint serves two kinds of consumers:

This page covers the infrastructure side: hosting, ports, identity, authentication, and operations. What users can do with the endpoint, and how to configure clients, lives in the AI documentation; the table Core Services vs AI documentation below shows where to go for what.

Understanding the MCP Endpoint Architecture

The MCP Endpoint is an ASP.NET Core web site that the installer creates in IIS, next to the Web Client, Web API, and Log API, on its own port. It answers MCP requests over Streamable HTTP at /mcp and reads its data from the Configuration Database.

graph LR subgraph clients ["MCP consumers"] Assistant["fa:fa-robot AI Assistant
in the Web Client"] External["fa:fa-laptop-code External MCP client
VS Code, Claude Desktop, Cline"] Local["fa:fa-terminal Local stdio client
NuGet package via dnx"] end subgraph core ["Nodinite Core Services (IIS)"] Mcp["fa:fa-plug MCP Endpoint
https://host:port/mcp"] end Idp["fab:fa-microsoft Identity provider
Microsoft Entra ID"] Config["fa:fa-database Configuration Database"] Assistant -->|"Server-to-server"| Mcp External -->|"Streamable HTTP + Bearer token"| Mcp External -.->|"Sign-in"| Idp Mcp -.->|"Token validation"| Idp Mcp -->|"Read, and write for Articles"| Config Local -->|"Own process, connection string"| Config style Mcp fill:#2678BD,color:#fff style Config fill:#3F827B,color:#fff style Idp fill:#9E59BB,color:#fff style Local fill:#E47528,color:#fff

Diagram: The IIS-hosted MCP Endpoint (blue) serves the AI Assistant and external MCP clients and reads the Configuration Database (green); the identity provider (purple) issues and validates tokens in OAuth mode. A local stdio client (orange) runs its own copy of the MCP Server from NuGet instead of calling the hosted endpoint.

Components

Component Role Requirements
MCP Endpoint (IIS web site) Hosts the MCP Server for the environment and answers MCP calls at /mcp A free port and a service identity, both set in the Portal; TLS strongly recommended
Configuration Database Source of the Repository Model, Monitor Views, Log Views, Articles, and other data the tools return The MCP Endpoint identity needs access, like the other Core Services
Identity provider (OAuth mode) Issues tokens to MCP clients; the endpoint validates audience and scopes App registration, audiences, and scopes; see MCP Endpoint External Access
AI Assistant Chat in the Web Client that uses the MCP tools EnableMcpChat in the AI System Parameter and an AI provider
External MCP client Calls the endpoint over HTTPS, or runs the NuGet package locally over stdio Client configuration from Set up AI Assistant and MCP Endpoint

Hosting and Authentication Modes

The MCP Server chooses one authentication mode when it starts:

Mode Used when Suitable for
OAuth 2.0 / JWT Bearer An identity provider discovery endpoint is configured ✅ Remote MCP clients and the AI Assistant in OAuth installations; the only mode that supports the standard MCP sign-in handshake
Windows (Negotiate) No identity provider is configured ⚠️ Same-domain callers only; standard remote MCP clients can't complete the MCP OAuth handshake
None (stdio) The NuGet package is started with --stdio Local development; the client process connects to the Configuration Database directly

Read Authentication Model for the discovery flow (/.well-known/oauth-protected-resource) and why remote clients need OAuth mode.

What You Set in the Nodinite Portal

You don't create the IIS site, bindings, or app pool by hand. You enter the values in the Nodinite Portal, and the generated installation package creates or updates the MCP Endpoint together with the other Core Services.

Setting Portal step What to enter
Port and service identity of the MCP Server Servers → Core platform services A free TCP port, approved for inbound traffic; the shared Core Services account or a dedicated identity
TLS host name and certificate subject Servers → Core host Configuration The FQDN clients use; it must match the certificate and the canonical MCP URL
Authentication mode Auth → Auth mode Windows Authentication or OAuth 2.0 / OIDC for the whole environment
MCP Server Audiences and Web Client MCP Server Scope (OAuth) Auth → MCP Server The MCP API app registration values, so the endpoint accepts tokens and the AI Assistant can request them
Keep the MCP Server stopped Servers → Core platform services Select it to deploy the MCP Endpoint without starting it, for example while you don't use AI features yet
Connection to the Configuration Database Infrastructure → SQL Server connectivity The connection string, shared by all Core Services

Before you download, check the assigned port under Review and Download → Allocated ports. For remote access, also align the canonical URL (Mcp:ResourceUrl) and token settings as described in MCP Endpoint External Access.

Important

From Nodinite 7.4.0, every installation and every update from an older version needs an available MCP Server port in the Portal. Plan the port and its firewall rule in the same change as the update.

System Parameters

The MCP Endpoint has no System Parameters of its own. One System Parameter changes what the AI features around it can do:

System Parameter Effect on the MCP Endpoint
AI → EnableAiForBizTalk Turns on the BizTalk tools; see BizTalk and MCP Endpoint
AI → EnableMcpChat Shows the AI Assistant chat in the Web Client; the chat uses the MCP Endpoint. External MCP clients don't need this flag

Changes take effect within about 15 seconds, without a restart; see Manage System Parameters.

What the Endpoint Exposes

The MCP tools are grouped by feature area: Systems, Services, Integrations, Endpoints, Message Types, Contracts, Domains, BPMs, Custom Metadata, Resources, Monitor Views, Monitoring Agents, Log Views, Articles, Snippets, Templates, and Document Profiles. Most tools read; the Articles, Snippets, Templates, and Document Profiles tools can also create and update. In OAuth mode, scopes such as mcp:repository.read limit what a client may call.

Core Services vs AI Documentation

Area Core Services documentation (this section) AI documentation
Focus Hosting, ports, identity, TLS, firewall, install, update, uninstall Benefits, usage, client setup, scopes, AI provider configuration
Typical reader Platform engineer, infrastructure administrator Solution architect, integration analyst, developer
Primary pages Prerequisites, Install MCP Endpoint, Update MCP Endpoint, MCP Endpoint Troubleshooting AI Overview, Set up AI Assistant and MCP Endpoint, MCP Endpoint for external AI tools, MCP Endpoint External Access, MCP Tool Reference

Connect Visual Studio Code

Visual Studio Code connects in one of two ways:

Frequently Asked Questions

Is the MCP client setup in the AI pages enough for production?

No. The AI pages explain client setup and usage. A production environment also needs the hosting described here: a port and identity in the Portal, TLS, firewall rules, and OAuth for remote clients.

Do I need this when updating from an older version?

Yes. When you update an environment from a version before 7.4.0, enter an MCP Server port in the Portal before you download the update package; see Update MCP Endpoint.

Does the AI Assistant need an external MCP client setup?

No. The AI Assistant calls the hosted endpoint from the Web Client; it needs EnableMcpChat and an AI provider. In OAuth installations, the Web Client also needs consent for the MCP scopes; see Connect the Web Client AI Assistant.

Does the MCP Endpoint send data to an AI provider?

No. The endpoint returns data to the MCP client that called it. Whatever that client, or the AI Assistant, sends to its language model is governed by that client's AI provider settings; see Security & Governance in AI Assistant.

Next Steps