Challenge 14 - Create the Task Tracking Agent

October 1, 2025 · View on GitHub

< Previous Challenge - Home - Next Challenge >

Challenge 14 - Create the Task Tracking Agent

Estimated Time: 30-45 minutes Difficulty: Intermediate Prerequisites: Complete Challenge 13 - Build Your Own MCP Server

Introduction

In this challenge, you will create the Task Tracking Agent, a specialized agent responsible for monitoring and managing the state of your multi-agent workflow. This agent will act as the central coordinator, ensuring that tasks are executed in the correct order and that the workflow progresses smoothly from start to finish.

You will define the agent's instructions, provide it with the necessary tools for state management, and integrate it into your existing multi-agent system.

Learning Objectives

By completing this challenge, you will understand:

  • How to create a specialized agent with a specific role and set of instructions
  • How to connect an agent to an MCP server to extend its capabilities
  • How to register MCP tools as plugins for a Semantic Kernel agent
  • The role of a task tracking agent in a multi-agent system

Architecture Context

The Task Tracking Agent is a critical component of your multi-agent system. It communicates with the MCP server you built in Challenge 13 to persist and retrieve workflow state, allowing it to coordinate the other agents in the system.

flowchart TD
    subgraph AgentLayer["Multi-Agent System"]
        Agent1["Agent #1<br/>(Data Import)"]
        Agent2["Agent #2<br/>(Market Analyst)"]
        Agent3["Agent #3<br/>(Content Creator)"]
        Agent4["Agent #4<br/>(Database Specialist)"]
        Agent5["Agent #5<br/>(Marketing)"]
        TaskAgent["Task Tracking Agent<br/>(MCP Client)"]
    end
    
    subgraph MCPLayer["MCP Server Layer"]
        MCPServer["MCP Server<br/>(State Management)"]
        
        subgraph Tools["MCP Tools"]
            WorkflowMgmt["Workflow Management"]
            StatePersist["State Persistence"]
            ProgressTrack["Progress Tracking"]
            DataStorage["Data Storage"]
        end
    end
    
    subgraph StateLayer["State Storage"]
        MemCache["IMemoryCache<br/>(In-Memory State)"]
        WorkflowData["Workflow Data<br/>• Step Progress<br/>• Agent Results<br/>• User Approvals"]
    end

    %% Agent interactions
    Agent1 -.-> TaskAgent
    Agent2 -.-> TaskAgent
    Agent3 -.-> TaskAgent
    Agent4 -.-> TaskAgent
    Agent5 -.-> TaskAgent
    
    %% MCP communication
    TaskAgent <-->|MCP Protocol| MCPServer
    
    %% Tool access
    MCPServer --> WorkflowMgmt
    MCPServer --> StatePersist
    MCPServer --> ProgressTrack
    MCPServer --> DataStorage
    
    %% State storage
    Tools --> MemCache
    MemCache --> WorkflowData

    %% Styling
    classDef agent fill:#e1f5fe,stroke:#0277bd,stroke-width:2px
    classDef mcp fill:#f3e5f5,stroke:#7b1fa2,stroke-width:2px
    classDef storage fill:#e8f5e8,stroke:#2e7d32,stroke-width:2px
    classDef tool fill:#fff3e0,stroke:#ef6c00,stroke-width:2px

    class Agent1,Agent2,Agent3,Agent4,Agent5,TaskAgent agent
    class MCPServer mcp
    class MemCache,WorkflowData storage
    class WorkflowMgmt,StatePersist,ProgressTrack,DataStorage tool

Challenges

1. Implement the Task Tracking Agent

In the AgentRegistry.cs file, you will implement the CreateTaskTrackingAgentAsync method. This method should perform the following steps:

  1. Clone the Kernel: Create a new kernel instance for the agent by cloning the one passed into the method. This ensures that each agent has its own isolated set of plugins.
  2. Connect to the MCP Server: Use the McpClientFactory to create a client that connects to the MCP server. The server endpoint is available in the IConfiguration service.
  3. List and Register MCP Tools: Once connected, list the available tools from the MCP server and register them as plugins with the agent's kernel. This will allow the agent to use the state management tools provided by the server.
  4. Define the Agent: Create a new ChatCompletionAgent with the following properties:
    • Name: Use the TaskTrackingAgentName constant.
    • Description: Provide a clear description of the agent's role. Refer to the Agent Specifications for the exact text.
    • Instructions: Provide detailed instructions for the agent. Refer to the Agent Specifications for the exact text.
    • Kernel: Assign the kernel you created in the first step.
  5. Add the Agent to the Registry: Add the newly created agent to the _agents dictionary.

Refer to the Coach/Dotnet/eShop/src/Agents/MultiAgent/AgentRegistry.cs file for a complete example of how to implement this method.

Success Criteria

✅ Task Tracking Agent Created: The CreateTaskTrackingAgentAsync method is fully implemented. ✅ MCP Connection Established: The agent successfully connects to the MCP server. ✅ MCP Tools Registered: The state management tools from the MCP server are registered as plugins for the agent. ✅ Agent Added to Registry: The Task Tracking Agent is added to the _agents dictionary and is available for use in the multi-agent system.

Resources

Next Steps

In Challenge 15, you will enhance your multi-agent system by implementing a custom InteractiveGroupChatManager that replaces the basic RoundRobin orchestrator.

< Previous Challenge - Home - Next Challenge >