Set up AI Assistant, Diagram Assistance, and MCP Endpoint in Nodinite
Since Nodinite 7.4.0, the AI Assistant, optional AI guidance in Diagram Assistance, BizTalk-specific AI workflows, and the MCP Endpoint are available to customers. This guide separates the setup paths so users can quickly see which steps apply to the integrated Web Client chat, which steps apply to Architecture Diagrams and C4 Diagrams, which steps apply to BizTalk-specific AI workflows, and which steps apply to a package-based MCP client such as VS Code.
Understanding the Setup Tracks
Integrated Web Client chat"] Start --> Diagram[" Diagram Assistance
Architecture Diagrams"] Start --> Mcp[" MCP Endpoint
External MCP client"] Chat --> ChatFlag[" AI System Parameter
EnableMcpChat = true"] Diagram --> DiagramFlag[" AI System Parameter
EnableDiagramAiAssistance = true"] ChatFlag --> Provider[" AI Assistant settings
Provider, Model Name, Endpoint, API Key"] DiagramFlag --> Provider Provider --> WebClient[" Ask questions in the Web Client"] Provider --> DiagramTab[" AI Assistance in Diagram Assistance"] Mcp --> McpJson[" .vscode/mcp.json"] McpJson --> Dnx[" dnx Nodinite.Core.McpServer"] Dnx --> Conn[" ConnectionStrings__NodiniteConfig"]
Diagram: The Web Client chat and the Diagram Assistance AI tab both depend on AI system parameter flags plus the shared provider settings. External MCP clients use .vscode/mcp.json, dnx, Nodinite.Core.McpServer, and ConnectionStrings__NodiniteConfig. BizTalk-specific AI workflows add EnableAiForBizTalk on top of the relevant client or administrator flow.
Important
Use the steps for the scenario you actually need. The AI Assistant and Diagram Assistance setup steps are performed in the Nodinite Web Client. The package-based MCP Endpoint setup is performed in
.vscode/mcp.jsonor the equivalent MCP configuration file used by your editor or client.
What You Need Before You Start
| Scenario | What you need | Where to configure |
|---|---|---|
| AI Assistant in the Web Client | Administrator Role, AI System Parameter, Provider, Model Name, Endpoint, API Key | AI System Parameter and AI Assistant |
| AI guidance in Diagram Assistance | Architecture Diagrams enabled, AI System Parameter, Provider, Model Name, Endpoint, API Key | AI System Parameter, AI Assistant, and Diagram Assistance |
| MCP Endpoint in VS Code or another MCP client | MCP-compatible client, .vscode/mcp.json, dnx, NuGet source, ConnectionStrings__NodiniteConfig |
MCP Endpoint |
| BizTalk-specific AI workflows | EnableAiForBizTalk, Administrator Role, and either external MCP setup or the BizTalk C4 generation path |
AI System Parameter, MCP Endpoint, and BizTalk and MCP Endpoint |
The MCP Endpoint track is always separate from the in-product AI flags. Use it whenever the user works from an external MCP client. If the goal is BizTalk analysis or BizTalk C4 generation, enable EnableAiForBizTalk as an additional gate.
Track 1 - Set up the AI Assistant in the Web Client
Step 1: Enable the AI feature
Open the AI System Parameter and enable the Web Client chat feature for the environment by setting EnableMcpChat to true.
{"EnableMcpChat":true,"EnableDiagramAiAssistance":false,"EnableAiForBizTalk":false}
This step enables the AI chat capability in the Web Client.
Step 2: Configure the AI provider from the AI Assistant settings
Open Administration → Settings → AI Assistant and enter the values for Provider, Model Name, Endpoint, and API Key.
| Setting | What to enter | Example values |
|---|---|---|
| Provider | The AI service you want Nodinite to use | OLLAMA, Azure OpenAI, OpenAI, GitHub Models |
| Model Name | The provider-specific model or deployment name | phi3:mini, gpt-4o, gpt-4o-mini |
| Endpoint | The provider base URL | http://localhost:11434, https://your-resource.openai.azure.com/, https://models.inference.ai.azure.com |
| API Key | The secret or token required by the selected Provider | Your provider API key or token |

Example of the AI Assistant settings dialog where you enter Provider, Model Name, Endpoint, and API Key before saving the configuration.
Step 3: Validate the integrated AI Assistant
Open the AI Assistant in the Web Client and run a simple query before you connect an external MCP client.
Validation Examples
- Show me all Services
- List integrations with gold level SLA
- Show me all Endpoints in the Finance domain
Expected Result
The AI Assistant opens, accepts the question, and returns data from the Nodinite Repository Model instead of a configuration or connection error.
Track 2 - Enable AI Assistance in Architecture Diagrams
Diagram Assistance is the set-aware guidance workspace for Architecture Diagrams and C4 Diagrams. Its Built-in Suggestions remain available without AI. Turn on EnableDiagramAiAssistance only when you also want the AI Assistance tab.
Step 1: Add the diagram AI flag to the AI system parameter
If you already enabled the integrated AI chat, update the JSON value to include both flags.
{"EnableMcpChat":true,"EnableDiagramAiAssistance":true,"EnableAiForBizTalk":false}
Step 2: Reuse the AI provider settings
Diagram Assistance uses the same Provider, Model Name, Endpoint, and API Key configured in Administration → Settings → AI Assistant. There is no separate provider dialog for C4 Diagrams.
Step 3: Validate the AI Assistance tab
Open Architecture Diagrams, select a Diagram Set, and open Diagram Assistance. Confirm that:
- Built-in Suggestions are visible for Add from Set, Coverage, and Review Impacts
- The AI Assistance tab is visible after
EnableDiagramAiAssistanceis enabled and the provider is configured
Track 3 - Set up the MCP Endpoint in VS Code or another MCP client
Use this track whenever the user works from an external MCP client. The same MCP setup is required whether the goal is Repository discovery, Architecture Diagrams, Diagram Sets, Mermaid C4, or BizTalk workflows. For BizTalk-specific tools, the environment must also have EnableAiForBizTalk enabled.
Step 1: Create or edit .vscode/mcp.json
If your workspace does not already contain .vscode/mcp.json, create it. If it already exists, merge the nodinite-mcp entry into the existing servers object instead of replacing other servers.
Step 2: Add the nodinite-mcp server configuration
{
"servers": {
"nodinite-mcp": {
"type": "stdio",
"command": "dnx",
"args": [
"Nodinite.Core.McpServer@7.4.50",
"--yes",
"--source", "https://your-package-feed.example/v3/index.json",
"--",
"--stdio"
],
"env": {
"ConnectionStrings__NodiniteConfig": "Server=.;Database=Nodinite_Config;Trusted_Connection=true;TrustServerCertificate=true"
}
}
}
}
Important
Nodinite.Core.McpServeris a .NET NuGet package. Because Nodinite ships monthly patches, review and update the package version frequently instead of leaving an older build pinned for long periods.
Step 3: Tune the values for your environment
The supplied example is not a universal copy-paste value. Tune these fields for your environment before you rely on the MCP client:
| Value | What to tune |
|---|---|
--source |
Point to the package feed you actually use |
ConnectionStrings__NodiniteConfig |
Point to your real Nodinite_Config database |
Nodinite.Core.McpServer@x.y.z |
Update the package version frequently. Since this is a .NET package and Nodinite ships monthly patches, move to the latest supported patch build regularly. |
Step 4: Reload the client and validate the MCP tools
Restart or reload your MCP-compatible client so it reads the updated .vscode/mcp.json file.
Then run a simple repository question such as:
- Show me all Applications in Production
- List integrations with gold level SLA
- Show me all Endpoints in the Finance domain
- List the send ports in the OrderProcessing BizTalk Application
BizTalk-specific addition to Track 3
If you plan to use BizTalk-specific MCP tools, enable EnableAiForBizTalk in the AI System Parameter.
{"EnableMcpChat":false,"EnableDiagramAiAssistance":false,"EnableAiForBizTalk":true}
This flag does not replace external MCP client setup. It adds the BizTalk-specific capability gate on top of the normal MCP connection model.
Current Delivery Model
The current delivery model now has three clear entry points:
- The integrated AI Assistant is configured from the Nodinite Web Client.
- Diagram Assistance adds built-in guidance to Architecture Diagrams and can add an AI Assistance tab when
EnableDiagramAiAssistanceis enabled. - The MCP Endpoint always requires MCP client registration and can then be consumed from an MCP-compatible client through the package-based
mcp.jsonsetup shown above. - BizTalk-specific AI workflows additionally require
EnableAiForBizTalk; external BizTalk MCP questions still use the same MCP client registration. - Both capabilities were released as part of Nodinite 7.4.0.
Troubleshooting Checks
The AI Assistant does not appear in the Web Client
Check the AI System Parameter first. If EnableMcpChat is still false, the AI Assistant will not appear.
The AI Assistant opens but cannot answer questions
Recheck Provider, Model Name, Endpoint, and API Key in the AI Assistant settings. A missing or invalid value in any of those fields prevents successful calls to the selected AI service.
The AI Assistance tab does not appear in Architecture Diagrams
Verify that EnableDiagramAiAssistance is set to true, the provider configuration is complete, and the environment has access to Architecture Diagrams. The built-in part of Diagram Assistance can still appear even when AI is disabled.
An external MCP client cannot connect
Verify that dnx is available, the NuGet --source is correct, and ConnectionStrings__NodiniteConfig points to the right Nodinite_Config database. If the package loads but the tools fail, the issue is usually in the client-side package or database connection values.
Repository questions work but BizTalk tools do not
Verify that EnableAiForBizTalk is enabled, that the user has Administrator access, and that the environment contains active BizTalk data. The MCP client registration alone is not enough for BizTalk-specific workflows.
The client loads but no Nodinite tools appear
Reload the client after editing .vscode/mcp.json and verify that the nodinite-mcp entry was merged into the existing servers object instead of being placed outside it.