Test Data Management

What is an MCP Server?

The Model Context Protocol (MCP) is an emerging open standard designed to improve how AI agents and applications interact with external tools and contextual data sources. Its primary goal is to enhance prompt generation and provide LLMs with more relevant, structured context.

An MCP Server can expose different types of functionality - ranging from simply delivering data to performing more complex actions. You can find available MCP Servers in the Official MCP Registry.

Connecting Data and Examine It in a chat

MCP Servers designed for data connectivity typically fall into two categories, the ones which can provide (read) data and those who support full CRUD operations.  Read-only MCP Servers guarantee that data cannot be changed, manipulated, or deleted. This is why the ODS AIConnect MCP Server is read-only.

Once you’ve registered an MCP Server with your AI agent (e.g., Claude Desktop or Microsoft Copilot for VS Code), you can begin by asking questions about your data in natural language.

⚠️ Although AI agents understand multiple languages, best results are typically achieved using English. 

Creating Your Own Analysis Tools

In programming environments such as Microsoft VS Code, an MCP Server can be orchestrated together with capabilities provided by Copilot – such as automated code generation. 

This makes it possible to:

• Develop custom analysis tools tailored to your specific data and use case.
• Iterate quickly using conversational prompts and AI-generated code.

Using ODS AI Connect

ODS AIConnect is an MCP Server specially designed to work with ASAM ODS servers in combination with Peak ASAM ODSBox.

Below is an example conversation flow:

1) Establish a Session 
Prompt: “Connect to my ASAM ODS server”
The agent requests the necessary login information, connects to the ODS server, and retrieves the data model (ontology) for Copilot.
 
 2) Ask Questions About the Data 
Prompt: “What kind of data is in the server?”
The agent returns an overview of the data. Copilot coordinates multiple MCP interactions to collect the required information. 

3) Start Analyzing the Data 

Prompt: “Compare tests of a certain campaign” 
Copilot generates and executes Python code against the data using ASAM ODSBox supported by the MCP Server. If errors occur, you can ask Copilot to fix them or refine the analysis. 

4) Create an Application
Prompt: “Transform this into a web app”
Once the analysis works as desired, you can ask Copilot to generate a web application.
In the example, the app was extended to allow switching between campaigns, updating the corresponding visualizations.

⚠️ Efficiency Tip

To optimize cost and performance, send complete requirements in a single request (e.g., a requirements.md document). You can further guide the agent using additional files such as agent.md and skill.md to obtain more accurate results.