Skip to main content
BuyLog In
WinForms Controls
Search in…
Docs > AI-powered Extensions
All docs
V26.1
    • WinForms Controls
    .NET Support and Migration
    System Requirements
    What's New
    What's Installed
    Build an Application
    Redistribution and Deployment
    Controls and Libraries
    Icon and Image Library
    AI-powered Extensions
    AI Assistant
    AI Chat Control
    Explain Formula
    Generate Image Description
    Prompt to Expression
    Smart Autocomplete
    Smart Paste
    Smart Search
    Semantic Search
    Custom AI-powered Extensions
    Common Features
    Security Considerations
    Data Access Security
    Hybrid & Enterprise Development
    Diagnostics and Troubleshooting
    UI Templates
    Desktop & Web App UI Framework (Powered by EF Core)
    Get More Help
    API Reference
    Download CHM

    DevExpress AI-powered Extensions for WinForms

    • 18 minutes to read

    Supported AI Clients

    Prerequisites

    DevExpress.AIIntegration assemblies reference the following versions of Microsoft.Extensions.AI.* NuGet packages:

    Package Name v26.1
    Microsoft.Extensions.AI 9.7.1
    Microsoft.Extensions.AI.OpenAI 9.7.1-preview.1.25365.4

    See the following breaking change advisory for more information: DevExpress.AIIntegration references stable versions of Microsoft AI packages.

    Install DevExpress NuGet Packages

    Install the following DevExpress NuGet Packages:

    1. DevExpress.AIIntegration.WinForms
    2. DevExpress.AIIntegration.OpenAI (required if you use the WinForms Chat control with OpenAI.Assistant)
    3. DevExpress.Win.Design (enables design-time features for DevExpress UI controls)

    Register AI Clients

    DevExpress AI-powered extensions operate within an AIExtensionsContainerDesktop container. This container manages all registered AI clients so that DevExpress UI controls can automatically leverage AI services.

    Register OpenAI Client

    The following code snippet registers an OpenAI client:

    using Microsoft.Extensions.AI;
using DevExpress.AIIntegration;

internal static class Program {
    [STAThread]
    static void Main() {
        Application.EnableVisualStyles();
        Application.SetCompatibleTextRenderingDefault(false);
        IChatClient openAIChatClient = new OpenAI.OpenAIClient(OpenAIKey).GetChatClient(Model)
            .AsIChatClient();
        AIExtensionsContainerDesktop.Default.RegisterChatClient(openAIChatClient);
        // Uncomment the following line if your project targets the .NET Framework and
        // you create AI-powered behaviors in code.
        // DevExpress.AIIntegration.WinForms.BehaviorInitializer.Initialize();
        Application.Run(new Form1());
    }
    static string OpenAIKey { get { return Environment.GetEnvironmentVariable("OPENAI_API_KEY"); } }
    static string Model { get { return "gpt-4o-mini"; } }
}

    Register Azure OpenAI Client

    The following code snippet registers an Azure OpenAI client:

    using Microsoft.Extensions.AI;
using DevExpress.AIIntegration;

internal static class Program {
    [STAThread]
    static void Main() {
        Application.EnableVisualStyles();
        Application.SetCompatibleTextRenderingDefault(false);
        IChatClient azureChatClient = new Azure.AI.OpenAI.AzureOpenAIClient(new Uri(AzureOpenAIEndpoint),
        new System.ClientModel.ApiKeyCredential(AzureOpenAIKey))
        .GetChatClient(ModelId).AsIChatClient();

        AIExtensionsContainerDesktop.Default.RegisterChatClient(azureChatClient);
        // Uncomment the following line if your project targets the .NET Framework and
        // you create AI-powered behaviors in code.
        // DevExpress.AIIntegration.WinForms.BehaviorInitializer.Initialize();
        Application.Run(new Form1());
    }
    static string AzureOpenAIEndpoint { get { return Environment.GetEnvironmentVariable("AZURE_OPENAI_ENDPOINT"); } }
    static string AzureOpenAIKey { get { return Environment.GetEnvironmentVariable("AZURE_OPENAI_APIKEY"); } }
    static string ModelId { get { return "gpt-4o-mini"; } }
}

    Register Anthropic Client

    Install the official Anthropic .NET SDK NuGet package to register a Claude chat client. The AsIChatClient method returns a Microsoft.Extensions.AI-compatible IChatClient. Pass the model name to this method.

    The AnthropicClient reads the API key from the ANTHROPIC_API_KEY environment variable, so no additional configuration is required when this variable is set.

    The following code snippet registers an Anthropic (Claude) client:

    using Anthropic;
using Microsoft.Extensions.AI;
using DevExpress.AIIntegration;

internal static class Program {
    [STAThread]
    static void Main() {
        Application.EnableVisualStyles();
        Application.SetCompatibleTextRenderingDefault(false);
        // The API key is read from the ANTHROPIC_API_KEY environment variable.
        // Pass the model name to AsIChatClient.
        IChatClient anthropicChatClient = new AnthropicClient().AsIChatClient("claude-haiku-4-5");
        AIExtensionsContainerDesktop.Default.RegisterChatClient(anthropicChatClient);
        // Uncomment the following line if your project targets the .NET Framework and
        // you create AI-powered behaviors in code.
        // DevExpress.AIIntegration.WinForms.BehaviorInitializer.Initialize();
        Application.Run(new Form1());
    }
}

    Enable Reasoning (Extended Thinking)

    To enable Claude’s extended thinking, configure chat client options and pass the thinking parameter through AdditionalProperties:

    using Anthropic;
using Microsoft.Extensions.AI;

IChatClient chatClient = new AnthropicClient()
    .AsIChatClient("claude-haiku-4-5")
    .AsBuilder()
    .ConfigureOptions(x => {
        x.Reasoning = new ReasoningOptions() {
            Effort = ReasoningEffort.Medium,
            Output = ReasoningOutput.Full
        };
        x.MaxOutputTokens = 4096;
        x.AdditionalProperties = new() {
            {
                "thinking", new Dictionary<string, object> {
                    { "type", "enabled" },
                    { "budget_tokens", 2048 } // The minimum allowed budget is usually 1024 or 2048 tokens.
                }
            }
        };
    })
    .Build();
AIExtensionsContainerDesktop.Default.RegisterChatClient(chatClient);

    Note

    Anthropic limitations

    Using Anthropic models with DevExpress AI-powered extensions has the following limitations:

    • No embedding model. Anthropic does not offer an embeddings API (source). Extensions and scenarios that rely on a vector store (for example, Semantic Search) require a separate, third-party embedding client (such as Voyage AI).
    • No image generation. Anthropic models cannot generate images. Extensions that produce images are not supported. Vision-based extensions that analyze existing images (such as Generate Image Description) continue to work.

    Register Semantic Kernel

    Install the connector package for the AI service. This example uses Microsoft.SemanticKernel.Connectors.Google.

    using Microsoft.Extensions.AI;
using Microsoft.SemanticKernel;
using Microsoft.SemanticKernel.ChatCompletion;
using Microsoft.SemanticKernel.Connectors.Google;
using DevExpress.AIIntegration;

internal static class Program {
    [STAThread]
    static void Main() {
        Application.EnableVisualStyles();
        Application.SetCompatibleTextRenderingDefault(false);
        var builder = Kernel.CreateBuilder().AddGoogleAIGeminiChatCompletion("YOUR_MODEL_ID", "YOUR_API_KEY", GoogleAIVersion.V1_Beta);
        Kernel kernel = builder.Build();
        IChatClient googleChatClient = kernel.GetRequiredService<IChatCompletionService>().AsChatClient();
        AIExtensionsContainerDesktop.Default.RegisterChatClient(googleChatClient);
        // Uncomment the following line if your project targets the .NET Framework and
        // you create AI-powered behaviors in code.
        // DevExpress.AIIntegration.WinForms.BehaviorInitializer.Initialize();
        Application.Run(new Form1());
    }
}

    Register Ollama Client

    The following code snippet registers an Ollama client:

    using OllamaSharp;
using Microsoft.Extensions.AI;
using DevExpress.AIIntegration;

internal static class Program {
    [STAThread]
    static void Main() {
        Application.EnableVisualStyles();
        Application.SetCompatibleTextRenderingDefault(false);
        IChatClient asChatClient = new OllamaApiClient(new Uri("http://localhost:11434/"), "MODEL_NAME");
        AIExtensionsContainerDesktop.Default.RegisterChatClient(asChatClient);
        // Uncomment the following line if your project targets the .NET Framework and
        // you create AI-powered behaviors in code.
        // DevExpress.AIIntegration.WinForms.BehaviorInitializer.Initialize();
        Application.Run(new Form1());
    }
}

    Register Foundry Local

    The following code snippet registers a Foundry Local chat client. Foundry Local runs AI models directly on the local device. It does not require cloud services or API keys. The SDK downloads and manages models automatically and exposes an OpenAI-compatible interface.

    Note

    Users do not need to install the Foundry Local CLI. The SDK is self-contained and handles model downloads and execution independently.

    using DevExpress.AIIntegration;
using Microsoft.AI.Foundry.Local;
using Microsoft.Extensions.AI;
using Microsoft.Extensions.Logging;
using OpenAI;
using System;
using System.ClientModel;
using System.Threading.Tasks;
using System.Windows.Forms;

namespace DXApplication {
    internal static class Program {
        /// <summary>
        /// The main entry point for the application.
        /// </summary>
        [STAThread]
        internal static void Main() {
            // Create a LoggerFactory and configure console logging.
            using var loggerFactory = LoggerFactory.Create(builder => {
                builder
                    // Set log level based on environment variable, fallback to Information
                    .SetMinimumLevel(Microsoft.Extensions.Logging.LogLevel.Information)
                    .AddSimpleConsole(options => { options.SingleLine = true; options.TimestampFormat = "HH:mm:ss "; });
            });

            // Specify the model name.
            string modelName = "phi-4-mini";

            // Register a Foundry Local client.
            RegisterFoundryLocalClient(modelName, loggerFactory).GetAwaiter();

            Application.EnableVisualStyles();
            Application.SetCompatibleTextRenderingDefault(false);
            Application.Run(new Form1());
        }

        async static Task RegisterFoundryLocalClient(string modelAlias, ILoggerFactory loggerFactory) {
            // Initialize the Foundry Local manager.
            var config = new Configuration {
                AppName = "DevExpressAIApp",
                LogLevel = Microsoft.AI.Foundry.Local.LogLevel.Information,
                Web = new Configuration.WebService() {
                    Urls = "http://127.0.0.1:52495"
                }
            };

            // Create a named logger.
            var logger = loggerFactory.CreateLogger("FoundryLocal");

            await FoundryLocalManager.CreateAsync(config, logger, null);

            var manager = FoundryLocalManager.Instance;
            var catalog = await manager.GetCatalogAsync();

            // Get the model from the catalog by alias.
            var model = await catalog.GetModelAsync(modelAlias) ?? throw new Exception($"Model {modelAlias} not found");

            // Check whether the model is cached and download it if required.
            if (!await model.IsCachedAsync()) {
                await model.DownloadAsync(progress => {
                    if (progress >= 100f) Console.WriteLine();
                });
            }

            // Load the model.
            await model.LoadAsync();

            // Start the web service.
            await manager.StartWebServiceAsync();

            // OpenAIClient requires an ApiKeyCredential, but local Foundry server runs without credentials.
            ApiKeyCredential key = new ApiKeyCredential("key_not_required");

            // Create the OpenAI client that points to the Foundry Local web service.
            OpenAIClient client = new OpenAIClient(key, new OpenAIClientOptions {
                Endpoint = new Uri(config.Web.Urls + "/v1"),
            });

            // Get the chat client.
            IChatClient chatClient = client.GetChatClient(model.Id).AsIChatClient();
            AIExtensionsContainerDesktop.Default.RegisterChatClient(chatClient);

            // Cleanup on exit.
            Application.ApplicationExit += async (sender, e) => {
                try {
                    chatClient.Dispose();
                    await model.UnloadAsync();
                }
                catch (Exception unloadEx) {
                    logger.LogWarning(unloadEx, "Error during model unload/cleanup.");
                }
            };
        }
    }
}

    See the following resources for additional information:

    Register ONNX Runtime

    The following code snippet registers an ONNX Runtime GenAI chat client. ONNX Runtime runs optimized AI models directly on the local device. It does not require cloud services.

    Note

    ONNX models are not bundled with the SDK. You must download an ONNX model to the local machine before use. You can obtain ONNX models from Hugging Face or ONNX Model Zoo. You can also convert custom models with ONNX converter tools.

    using DevExpress.AIIntegration;
using Microsoft.Extensions.AI;
using Microsoft.ML.OnnxRuntimeGenAI;

// Define the path to the ONNX model directory.
var modelPath = "..\\..\\models\\cpu_and_mobile\\cpu-int4-rtn-block-32-acc-level-4\\";
// The path for Linux-based systems
// var modelPath = "../../models/cpu_and_mobile/cpu-int4-rtn-block-32-acc-level-4/";

// Verify that the model directory exists.
if (!Directory.Exists(modelPath)) {
    MessageBox.Show("Model files not found. Ensure that the model files exist at the specified path.");
    return;
}

// Create the model configuration.
using var config = new Config(modelPath);

// Initialize the ONNX model.
using var model = new Model(config);

// Create an ONNX Runtime chat client.
using var onnxChatClient = new OnnxRuntimeGenAIChatClient(model);

// Configure chat client options.
using var chat = onnxChatClient.AsBuilder().ConfigureOptions(
    x => x.MaxOutputTokens = 4096
).Build();

// Register the chat client.
AIExtensionsContainerDesktop.Default.RegisterChatClient(chat);

// Handle application exit and unload the model.
Application.ApplicationExit += (sender, e) => {
    chat?.Dispose();
    onnxChatClient?.Dispose();
    model?.Dispose();
    config?.Dispose();
};

    Tip

    • Set the model path to the directory containing ONNX model files (usually includes genai_config.json, model weights, and tokenizer files).
    • Configure MaxOutputTokens based on the model’s capabilities and your application’s requirements.
    • ONNX models may have different configuration requirements. Refer to the model’s documentation on Hugging Face or the model source.

    See the following resources for additional information:

    AI-powered Extensions

    AI Assistant (Text Transform)

    AI Assistant extensions allow you to enhance the way your users interact with and manage text content with AI-powered precision. These extensions leverage advanced natural language processing (NLP) technologies to provide automated, intelligent text manipulation capabilities directly within your Windows Forms applications.

    WinForms MemoEdit with AI-powered Options, DevExpress

    Run Demo

    AI-powered options include:

    • Change Style
    • Change Tone
    • Expand
    • Explain
    • Proofread
    • Shorten
    • Summarize
    • Translate
    • Ask AI Assistant (allows users to interact with an AI-powered assistant directly within your application)

    Applies to:

    See the following help topic for additional information: AI Assistant Extensions.

    Explain Formula

    The AI-powered “Explain Formula” extension generates a detailed explanation of the formula used in a worksheet cell in the DevExpress Spreadsheet control.

    Explain Formula - WinForms Spreadsheet, DevExpress

    Run Demo

    See the following help topic for additional information: Explain Formula.

    Generate Image Description

    The AI-powered “Generate Image Description” extension generates the description for the image in DevExpress WinForms Spreadsheet and Rich Text Edit controls.

    Play the following animation to see how the AI-powered “Generate Image Description” extension in a WinForms Rich Text Editor generates Alt text for an image:

    AI-powered Generate Image Description Behavior - WinForms Rich Text Editor, DevExpress

    Run Demo

    See the following help topic for additional information: Generate Image Description.

    Prompt to Expression

    The AI-powered “Prompt to Expression” extension converts natural language into valid filter and unbound column expressions for data-aware WinForms controls. Instead of writing complex expressions, users describe the desired logic in plain text:

    Sample Filter Expression
    Display orders expected to arrive within 7 days.
    Sample Unbound Column Expression
    Compute total amount.

    The system sends the prompt to the configured AI service, which generates a valid expression for the control. The Expression Editor or Filter Editor displays and validates the result immediately.

    Prompt to Expression - WinForms UI Controls, DevExpress

    Run Demo: Prompt to Expression

    See the following help topic for more information: Prompt to Expression.

    Smart Autocomplete

    The AI-powered “Smart Autocomplete” feature intelligently predicts and suggests words or phrases based on the user’s current input. As you type, the AI model analyzes the context of the text and makes relevant suggestions in real time.

    Run Demo

    Applies to:

    MemoEdit

    See the following help topic for additional information: Smart Autocomplete.

    Smart Paste

    “SmartPaste” is an AI-powered feature that transforms the traditional copy-and-paste process into a smarter, more efficient tool. Designed to improve productivity, SmartPaste analyzes the content you copy and intelligently assigns the right values to the appropriate fields or row cells in the DevExpress Data Grid and LayoutControl-driven forms.

    Play the following animation to see how “SmartPaste” works:

    AI-powered Smart Paste Extension for WinForms | DevExpress

    Run Demo

    Applies to:

    See the following help topic for additional information: Smart Paste.

    “Smart Search” works alongside the traditional search algorithms to offer a more powerful and user-friendly search experience. It offers results that are more aligned with what the user is seeking, even if the input contains misspellings.

    Play the following animation to see how AI-powered smart search works in the DevExpress WinForms Ribbon control:

    Smart Search - WinForms Ribbon Control, DevExpress

    Run Demo

    Applies to:

    See the following help topic for additional information: Smart Search.

    Semantic search enables users to locate relevant data quickly and accurately within large datasets. Unlike standard keyword-based search, semantic search leverages Natural Language Processing (NLP) to analyze search queries beyond exact keyword matching.

    Semantic search uses an embedding generator to convert text into numerical vector representations. Vectors are stored in a vector database. When a user enters a search query, the search engine computes similarity scores between the query vector and stored data vectors to return the most relevant results.

    AI-driven Semantic Search - WinForms Data Grid, DevExpress

    Run Demo: Semantic Search - Grid Control

    Applies to:

    See the following help topic for additional information: Semantic Search.

    Custom Extensions

    You can create custom extensions based on DevExpress AI-powered extensions. See the following help topic for additional information and examples: Create Custom AI-powered Extensions.

    Custom AI-powered Extension - WinForms Rich Text Editor, DevExpress

    Run Demo

    Security Considerations (Indirect Prompt Injection)

    DevExpress text-based AI-powered extensions and the document-based Ask AI extension process user-provided content (documents, messages, or clipboard data) and pass this content to an LLM for analysis. This workflow introduces a class of attacks (indirect prompt injection) where malicious instructions are embedded in valid content. These instructions may attempt to override system behavior, extract sensitive information, or manipulate the model’s output.

    DevExpress AI-powered extensions include automatic prompt-injection protection. Protection is enabled by default for all text and document-based AI-powered extensions. The system adds protective instructions to AI requests. The LLM interprets these instructions and determines the final response.

    Refer to the following help topic for additional information: Prompt Injection Protection in AI-powered Extensions.

    AI Chat Control

    Note

    The AI Chat Control leverages BlazorWebView to reuse the DevExpress Blazor AI Chat component (DxAIChat). To use the WinForms AI Chat Control you must have one of the following active subscriptions: Universal, DXperience, ASP.NET & Blazor.

    The AI Chat Control (AIChatControl) allows you to incorporate an interactive, Copilot-inspired chat-based UI within your WinForms application. The DevExpress AI Chat Control (AIChatControl) can only be used in Windows Forms applications that target the .NET 8+ framework.

    Run Demo: AI Chat Control

    WinForms AI Chat Control, DevExpress

    Features include:

    • Integrate seamlessly with AI services
    • Render Markdown messages
    • Copy and regenerate responses
    • Manually handle chat messages
    • Attach files
    • Specify prompt suggestions
    • Create an Assistant that chats based on your data
    • Manage multiple chat clients or agents
    • Save and load chat history
    • Response streaming
    • Resources
    • Tool Calling
    • DevExpress Skins

    See the following help topic for additional information: AI Chat Control.

    See Also