Build Persistent AI Agents with .NET SDK

Azure.AI.Agents.Persistent (.NET)

Low-level SDK for creating and managing persistent AI agents with threads, messages, runs, and tools.

Installation

dotnet add package Azure.AI.Agents.Persistent --prerelease
dotnet add package Azure.Identity

Current Versions: Stable v1.1.0, Preview v1.2.0-beta.8

Environment Variables

PROJECT_ENDPOINT=https://<resource>.services.ai.azure.com/api/projects/<project>
MODEL_DEPLOYMENT_NAME=gpt-4o-mini
AZURE_BING_CONNECTION_ID=<bing-connection-resource-id>
AZURE_AI_SEARCH_CONNECTION_ID=<search-connection-resource-id>

Authentication

using Azure.AI.Agents.Persistent;
using Azure.Identity;

var projectEndpoint = Environment.GetEnvironmentVariable("PROJECT_ENDPOINT");
PersistentAgentsClient client = new(projectEndpoint, new DefaultAzureCredential());

Client Hierarchy

PersistentAgentsClient
├── Administration  → Agent CRUD operations
├── Threads         → Thread management
├── Messages        → Message operations
├── Runs            → Run execution and streaming
├── Files           → File upload/download
└── VectorStores    → Vector store management

Core Workflow

1. Create Agent

var modelDeploymentName = Environment.GetEnvironmentVariable("MODEL_DEPLOYMENT_NAME");

PersistentAgent agent = await client.Administration.CreateAgentAsync(
    model: modelDeploymentName,
    name: "Math Tutor",
    instructions: "You are a personal math tutor. Write and run code to answer math questions.",
    tools: [new CodeInterpreterToolDefinition()]
);

2. Create Thread and Message

// Create thread
PersistentAgentThread thread = await client.Threads.CreateThreadAsync();

// Create message
await client.Messages.CreateMessageAsync(
    thread.Id,
    MessageRole.User,
    "I need to solve the equation `3x + 11 = 14`. Can you help me?"
);

3. Run Agent (Polling)

// Create run
ThreadRun run = await client.Runs.CreateRunAsync(
    thread.Id,
    agent.Id,
    additionalInstructions: "Please address the user as Jane Doe."
);

// Poll for completion
do
{
    await Task.Delay(TimeSpan.FromMilliseconds(500));
    run = await client.Runs.GetRunAsync(thread.Id, run.Id);
}
while (run.Status == RunStatus.Queued || run.Status == RunStatus.InProgress);

// Retrieve messages
await foreach (PersistentThreadMessage message in client.Messages.GetMessagesAsync(
    threadId: thread.Id, 
    order: ListSortOrder.Ascending))
{
    Console.Write($"{message.Role}: ");
    foreach (MessageContent content in message.ContentItems)
    {
        if (content is MessageTextContent textContent)
            Console.WriteLine(textContent.Text);
    }
}

4. Streaming Response

AsyncCollectionResult<StreamingUpdate> stream = client.Runs.CreateRunStreamingAsync(
    thread.Id, 
    agent.Id
);

await foreach (StreamingUpdate update in stream)
{
    if (update.UpdateKind == StreamingUpdateReason.RunCreated)
    {
        Console.WriteLine("--- Run started! ---");
    }
    else if (update is MessageContentUpdate contentUpdate)
    {
        Console.Write(contentUpdate.Text);
    }
    else if (update.UpdateKind == StreamingUpdateReason.RunCompleted)
    {
        Console.WriteLine("\n--- Run completed! ---");
    }
}

5. Function Calling

// Define function tool
FunctionToolDefinition weatherTool = new(
    name: "getCurrentWeather",
    description: "Gets the current weather at a location.",
    parameters: BinaryData.FromObjectAsJson(new
    {
        Type = "object",
        Properties = new
        {
            Location = new { Type = "string", Description = "City and state, e.g. San Francisco, CA" },
            Unit = new { Type = "string", Enum = new[] { "c", "f" } }
        },
        Required = new[] { "location" }
    }, new JsonSerializerOptions { PropertyNamingPolicy = JsonNamingPolicy.CamelCase })
);

// Create agent with function
PersistentAgent agent = await client.Administration.CreateAgentAsync(
    model: modelDeploymentName,
    name: "Weather Bot",
    instructions: "You are a weather bot.",
    tools: [weatherTool]
);

// Handle function calls during polling
do
{
    await Task.Delay(500);
    run = await client.Runs.GetRunAsync(thread.Id, run.Id);

    if (run.Status == RunStatus.RequiresAction 
        && run.RequiredAction is SubmitToolOutputsAction submitAction)
    {
        List<ToolOutput> outputs = [];
        foreach (RequiredToolCall toolCall in submitAction.ToolCalls)
        {
            if (toolCall is RequiredFunctionToolCall funcCall)
            {
                // Execute function and get result
                string result = ExecuteFunction(funcCall.Name, funcCall.Arguments);
                outputs.Add(new ToolOutput(toolCall, result));
            }
        }
        run = await client.Runs.SubmitToolOutputsToRunAsync(run, outputs, toolApprovals: null);
    }
}
while (run.Status == RunStatus.Queued || run.Status == RunStatus.InProgress);

6. File Search with Vector Store

// Upload file
PersistentAgentFileInfo file = await client.Files.UploadFileAsync(
    filePath: "document.txt",
    purpose: PersistentAgentFilePurpose.Agents
);

// Create vector store
PersistentAgentsVectorStore vectorStore = await client.VectorStores.CreateVectorStoreAsync(
    fileIds: [file.Id],
    name: "my_vector_store"
);

// Create file search resource
FileSearchToolResource fileSearchResource = new();
fileSearchResource.VectorStoreIds.Add(vectorStore.Id);

// Create agent with file search
PersistentAgent agent = await client.Administration.CreateAgentAsync(
    model: modelDeploymentName,
    name: "Document Assistant",
    instructions: "You help users find information in documents.",
    tools: [new FileSearchToolDefinition()],
    toolResources: new ToolResources { FileSearch = fileSearchResource }
);

7. Bing Grounding

var bingConnectionId = Environment.GetEnvironmentVariable("AZURE_BING_CONNECTION_ID");

BingGroundingToolDefinition bingTool = new(
    new BingGroundingSearchToolParameters(
        [new BingGroundingSearchConfiguration(bingConnectionId)]
    )
);

PersistentAgent agent = await client.Administration.CreateAgentAsync(
    model: modelDeploymentName,
    name: "Search Agent",
    instructions: "Use Bing to answer questions about current events.",
    tools: [bingTool]
);

8. Azure AI Search

AzureAISearchToolResource searchResource = new(
    connectionId: searchConnectionId,
    indexName: "my_index",
    topK: 5,
    filter: "category eq 'documentation'",
    queryType: AzureAISearchQueryType.Simple
);

PersistentAgent agent = await client.Administration.CreateAgentAsync(
    model: modelDeploymentName,
    name: "Search Agent",
    instructions: "Search the documentation index to answer questions.",
    tools: [new AzureAISearchToolDefinition()],
    toolResources: new ToolResources { AzureAISearch = searchResource }
);

9. Cleanup

await client.Threads.DeleteThreadAsync(thread.Id);
await client.Administration.DeleteAgentAsync(agent.Id);
await client.VectorStores.DeleteVectorStoreAsync(vectorStore.Id);
await client.Files.DeleteFileAsync(file.Id);

Available Tools

Tool	Class	Purpose
Code Interpreter	CodeInterpreterToolDefinition	Execute Python code, generate visualizations
File Search	FileSearchToolDefinition	Search uploaded files via vector stores
Function Calling	FunctionToolDefinition	Call custom functions
Bing Grounding	BingGroundingToolDefinition	Web search via Bing
Azure AI Search	AzureAISearchToolDefinition	Search Azure AI Search indexes
OpenAPI	OpenApiToolDefinition	Call external APIs via OpenAPI spec
Azure Functions	AzureFunctionToolDefinition	Invoke Azure Functions
MCP	MCPToolDefinition	Model Context Protocol tools
SharePoint	SharepointToolDefinition	Access SharePoint content
Microsoft Fabric	MicrosoftFabricToolDefinition	Access Fabric data

Streaming Update Types

Update Type	Description
StreamingUpdateReason.RunCreated	Run started
StreamingUpdateReason.RunInProgress	Run processing
StreamingUpdateReason.RunCompleted	Run finished
StreamingUpdateReason.RunFailed	Run errored
MessageContentUpdate	Text content chunk
RunStepUpdate	Step status change

Key Types Reference

Type	Purpose
PersistentAgentsClient	Main entry point
PersistentAgent	Agent with model, instructions, tools
PersistentAgentThread	Conversation thread
PersistentThreadMessage	Message in thread
ThreadRun	Execution of agent against thread
RunStatus	Queued, InProgress, RequiresAction, Completed, Failed
ToolResources	Combined tool resources
ToolOutput	Function call response

Best Practices

1. Always dispose clients - Use using statements or explicit disposal
2. Poll with appropriate delays - 500ms recommended between status checks
3. Clean up resources - Delete threads and agents when done
4. Handle all run statuses - Check for RequiresAction, Failed, Cancelled
5. Use streaming for real-time UX - Better user experience than polling
6. Store IDs not objects - Reference agents/threads by ID
7. Use async methods - All operations should be async

Error Handling

using Azure;

try
{
    var agent = await client.Administration.CreateAgentAsync(...);
}
catch (RequestFailedException ex) when (ex.Status == 404)
{
    Console.WriteLine("Resource not found");
}
catch (RequestFailedException ex)
{
    Console.WriteLine($"Error: {ex.Status} - {ex.ErrorCode}: {ex.Message}");
}

Related SDKs

SDK	Purpose	Install
Azure.AI.Agents.Persistent	Low-level agents (this SDK)	dotnet add package Azure.AI.Agents.Persistent
Azure.AI.Projects	High-level project client	dotnet add package Azure.AI.Projects

Reference Links

Resource	URL
NuGet Package	https://www.nuget.org/packages/Azure.AI.Agents.Persistent
API Reference	https://learn.microsoft.com/dotnet/api/azure.ai.agents.persistent
GitHub Source	https://github.com/Azure/azure-sdk-for-net/tree/main/sdk/ai/Azure.AI.Agents.Persistent
Samples	https://github.com/Azure/azure-sdk-for-net/tree/main/sdk/ai/Azure.AI.Agents.Persistent/samples

When to Use

This skill is applicable to execute the workflow or actions described in the overview.

Limitations

• Use this skill only when the task clearly matches the scope described above.
• Do not treat the output as a substitute for environment-specific validation, testing, or expert review.
• Stop and ask for clarification if required inputs, permissions, safety boundaries, or success criteria are missing.