{"id":4509,"date":"2025-03-17T18:22:32","date_gmt":"2025-03-18T01:22:32","guid":{"rendered":"https:\/\/devblogs.microsoft.com\/semantic-kernel\/?p=4509"},"modified":"2025-03-17T18:22:32","modified_gmt":"2025-03-18T01:22:32","slug":"guest-blog-build-a-multi-agent-system-using-microsoft-azure-ai-agent-service-and-semantic-kernel-in-3-simple-steps","status":"publish","type":"post","link":"https:\/\/devblogs.microsoft.com\/agent-framework\/guest-blog-build-a-multi-agent-system-using-microsoft-azure-ai-agent-service-and-semantic-kernel-in-3-simple-steps\/","title":{"rendered":"Guest Blog: Build a Multi-Agent System Using Microsoft Azure AI Agent Service and Semantic Kernel in 3 Simple Steps!"},"content":{"rendered":"

Build a Multi-Agent System Using Microsoft Azure AI Agent Service and Semantic Kernel in 3 Simple Steps!<\/h4>\n

<\/p>\n

Today we’re thrilled to welcome back guest author,\u00a0Akshay Kokane<\/a> to share his recent Medium article on Build a Multi-Agent System Using Microsoft Azure AI Agent Service and Semantic Kernel in 3 Simple Steps<\/a>. We\u2019ll turn it over to him to dive in!<\/p>\n

In my previous blog<\/a>, I introduced\u00a0Microsoft\u2019s Azure AI Agent Service<\/strong>, a fully managed platform that simplifies the process of\u00a0building, deploying, and scaling AI agents<\/strong>. Unlike OpenAI Assistant, Azure AI Agent Service offers greater flexibility<\/strong>, supporting multiple\u00a0LLMs<\/strong>\u00a0(OpenAI, Mistral, DeepSeek, etc.) while ensuring\u00a0enterprise-grade security<\/strong> and seamless integration with the Azure ecosystem<\/strong>.<\/p>\n

\n
\n
\n
That blog provided a step-by-step guide<\/strong>\u00a0on creating an AI agent using\u00a0Azure AI Foundry Hub<\/strong>, covering everything from\u00a0project setup<\/strong>\u00a0and\u00a0agent configuration<\/strong>\u00a0to\u00a0grounding it with external data<\/strong>\u00a0(e.g.,\u00a010-K reports<\/strong>) and integrating it into a\u00a0.NET application<\/strong>\u00a0for production use.<\/div>\n<\/div>\n<\/div>\n<\/div>\n

By combining\u00a0no-code agent creation<\/strong>\u00a0with\u00a0custom .NET integration<\/strong>, this approach enables developers to build\u00a0scalable and intelligent AI-driven assistants<\/strong>\u00a0with minimal effort, leveraging\u00a0structured data<\/strong>\u00a0and\u00a0robust security<\/strong>.<\/p>\n

In this blog, I will take the discussion further by demonstrating how to build a multi-agent system using\u00a0Microsoft\u2019s AI Framework \u2014 Semantic Kernel<\/strong>.<\/p>\n

Previously, we created\u00a0StockExpertAgent<\/strong>, an AI agent grounded with companies\u2019\u00a010-K reports<\/strong>. Now, we will introduce InvestorAgent, which will based in the portfolio and data from stock agent provide me with personalized advice. Using\u00a0Semantic Kernel<\/strong>, we will orchestrate these agents to enable\u00a0intelligent interactions and decision-making<\/strong>\u00a0in a\u00a0multi-agent framework<\/strong>.<\/p>\n

\n
\n
\"\"<\/picture><\/div>\n<\/div>
An image illustrating the key topics covered in this blog.<\/figcaption><\/figure>\n

Let\u2019s start!<\/p>\n

Step 1: Create InvestorAgent in Azure AI Agent Service<\/strong><\/h2>\n

For a detailed guide on creating agents in the Azure AI Agent Service, refer to my previous\u00a0blog<\/a>. The process is straightforward once you set up an AI Foundry Project and Hub. Simply navigate to\u00a0Azure AI Studio<\/a>\u00a0and select the\u00a0Agents<\/strong>\u00a0tab to get started.<\/p>\n

I am going to ground it up with with sample investment portfolio.<\/p>\n

\n
\n
\"\"<\/picture><\/div>\n<\/div>
Screenshot from Azure AI Studio Agents tab<\/figcaption><\/figure>\n

Step 2: Retrieve Agents defined in Azure AI Service in step 1<\/strong><\/h2>\n

Now in .NET app, install following nugets<\/p>\n

#r \"nuget: Azure.AI.Projects, 1.0.0-beta.4\"\r\n#r \"nuget: Azure.Identity\"\r\n#r \"nuget: Microsoft.SemanticKernel.Agents.Core, 1.40.0-preview\"\r\n#r \"nuget: Microsoft.SemanticKernel.Agents.AzureAI, 1.40.0-preview\"<\/code><\/pre>\n
\/\/ Create Kernel\r\n        var builder = Kernel.CreateBuilder();\r\n        builder.AddAzureOpenAIChatCompletion(\"GPT4ov1\", \"https:\/\/testmediumazureopenai.openai.azure.com\", \"<api-key>\");\r\n        Kernel kernel = builder.Build();\r\n\r\n        \/\/ Define the connection string to connect with Azure AI Agents\r\n        string connectionString = \"eastus2.api.azureml.ms;<project>;aifoundary-hub;aiproject\";\r\n        \r\n        \/\/ Initialize the Azure Agents Client\r\n        AgentsClient client = new AgentsClient(connectionString, credential);\r\n        \r\n        \/\/ Initialize the Stock Expert Agent\r\n        string stockExpertAgentId = \"asst_8Ev5OpD3bMUll20xJFu36Pah\";\r\n        Response<Agent> agentResponse = await client.GetAgentAsync(stockExpertAgentId);\r\n        Agent stockExpertAgentDefinition = agentResponse.Value;\r\n        AzureAIAgent stockExpertAgent = new(stockExpertAgentDefinition, client)\r\n        {\r\n            Kernel = kernel \/\/ Assigning kernel for execution\r\n        };<\/code><\/pre>\n
Now we will define Investor Advisor agent which has been grounded with sample portfolio<\/p>\n
\/\/ Initialize Inventor Advisor Proxy Agent\r\n            string investorAdvisorAgentId = \"asst_txX6WR6caXzpVhEksCupcySS\";\r\n            Response<Agent> investorAdvAgentResponse = await client.GetAgentAsync(investorAdvisorAgentId);\r\n            Agent investorAdvAgentDefinition = investorAdvAgentResponse.Value;\r\n            AzureAIAgent investorAdvAgent = new(investorAdvAgentDefinition, client)\r\n            {\r\n                Kernel = kernel,\r\n            };<\/code><\/pre>\n
Step 3: Define Agent Group Chat and see it in action<\/strong><\/h2>\n
Using\u00a0Semantic Kernel\u2019<\/strong>s Group Chat, create multi-agent system.<\/p>\n
        \/\/ Create Semantic Kernel Agent group chat with both agents\r\n        AgentGroupChat agentGroupChat = new AgentGroupChat(stockExpertAgent, investorProxyAgent)\r\n        {\r\n            ExecutionSettings = new()\r\n            {\r\n                SelectionStrategy = selectionStrategy,\r\n                TerminationStrategy = terminationStrategy,\r\n            }\r\n        };\r\n\r\n        \/\/ Start the interaction\r\n        agentGroupChat.AddChatMessage(new ChatMessageContent(AuthorRole.User, \"<Starting prompt here>\"));\r\n        \r\n        \/\/ Process and display conversation messages asynchronously\r\n        await foreach (var content in agentGroupChat.InvokeAsync())\r\n        {\r\n            string interaction = $\"<b>#{content.Role}<\/b> - <i>{content.AuthorName ?? \"*\"}<\/i>: \\\"{content.Content}\\\"\";\r\n            Console.WriteLine(interaction);\r\n        }<\/code><\/pre>\n
\n
\"\"<\/picture><\/div>\n<\/div>\n
Stock Expert Agent provided company insights for investment<\/p>\n
\n
\n
\"\"<\/picture><\/div>\n<\/div>
Then Investment Advisor agent went ahead and provided with the investment advice<\/figcaption><\/figure>\n
About how to define Termination and Selection Stratergy checkout my previous blog here<\/a>. You can find complete code here<\/a>.<\/p>\n
Conclusion<\/h1>\n
In the steps above, we demonstrated how AI Agents created in Azure AI Agent Service can be orchestrated using Semantic Kernel. The process leveraged a knowledge source built in Azure AI Foundry and performed Retrieval-Augmented Generation (RAG) as needed.<\/p>\n
However, after running the application, you may notice that the correct agent is not always selected. To address this, we need to define a termination and selection strategy for the agent group chat.<\/p>\n
Overall, I prefer using managed services, but I\u2019ve observed that the\u00a0ChatCompletionAgent<\/strong>\u00a0in\u00a0Semantic Kernel<\/strong>\u00a0has lower latency compared to\u00a0Azure Agent<\/strong>. This is expected since we are leveraging a managed service, which likely introduces some additional overhead.<\/p>\n
From the Semantic Kernel team, we\u2019d like to thank Akshay for his time and all of his great work. \u00a0Please reach out if you have any questions or feedback through our\u00a0Semantic Kernel GitHub Discussion Channel<\/a>. We look forward to hearing from you!<\/p>\n
 <\/p>\n
 <\/p>\n","protected":false},"excerpt":{"rendered":"
Build a Multi-Agent System Using Microsoft Azure AI Agent Service and Semantic Kernel in 3 Simple Steps! Today we’re thrilled to welcome back guest author,\u00a0Akshay Kokane to share his recent Medium article on Build a Multi-Agent System Using Microsoft Azure AI Agent Service and Semantic Kernel in 3 Simple Steps. We\u2019ll turn it over to […]<\/p>\n","protected":false},"author":149071,"featured_media":4512,"comment_status":"open","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"_acf_changed":false,"footnotes":""},"categories":[117],"tags":[48,127,63,9],"class_list":["post-4509","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-guest-blog","tag-ai","tag-azure-ai-agents","tag-microsoft-semantic-kernel","tag-semantic-kernel"],"acf":[],"blog_post_summary":"
Build a Multi-Agent System Using Microsoft Azure AI Agent Service and Semantic Kernel in 3 Simple Steps! Today we’re thrilled to welcome back guest author,\u00a0Akshay Kokane to share his recent Medium article on Build a Multi-Agent System Using Microsoft Azure AI Agent Service and Semantic Kernel in 3 Simple Steps. We\u2019ll turn it over to […]<\/p>\n","_links":{"self":[{"href":"https:\/\/devblogs.microsoft.com\/agent-framework\/wp-json\/wp\/v2\/posts\/4509","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/devblogs.microsoft.com\/agent-framework\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/devblogs.microsoft.com\/agent-framework\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/devblogs.microsoft.com\/agent-framework\/wp-json\/wp\/v2\/users\/149071"}],"replies":[{"embeddable":true,"href":"https:\/\/devblogs.microsoft.com\/agent-framework\/wp-json\/wp\/v2\/comments?post=4509"}],"version-history":[{"count":0,"href":"https:\/\/devblogs.microsoft.com\/agent-framework\/wp-json\/wp\/v2\/posts\/4509\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/devblogs.microsoft.com\/agent-framework\/wp-json\/wp\/v2\/media\/4512"}],"wp:attachment":[{"href":"https:\/\/devblogs.microsoft.com\/agent-framework\/wp-json\/wp\/v2\/media?parent=4509"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/devblogs.microsoft.com\/agent-framework\/wp-json\/wp\/v2\/categories?post=4509"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/devblogs.microsoft.com\/agent-framework\/wp-json\/wp\/v2\/tags?post=4509"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}