microsoft--agent-framework/dotnet/samples/02-agents/AgentProviders/github-copilot/Agent_With_GitHubCopilot

Prerequisites

⚠️ WARNING: Container Recommendation

GitHub Copilot can execute tools and commands that may interact with your system. For safety, it is strongly recommended to run this sample in a containerized environment (e.g., Docker, Dev Container) to avoid unintended consequences to your machine.

Before you begin, ensure you have the following prerequisites:

• .NET 10 SDK or later
• GitHub Copilot CLI installed and available in your PATH (or provide a custom path)

Setting up GitHub Copilot CLI

To use this sample, you need to have the GitHub Copilot CLI installed. You can install it by following the instructions at: https://github.com/github/copilot-sdk

Once installed, ensure the copilot command is available in your PATH, or configure a custom path using CopilotClientOptions.

Running the Sample

No additional environment variables are required if using default configuration. The sample will:

1. Create a GitHub Copilot client with default options
2. Create an AI agent using the Copilot SDK
3. Send a message to the agent
4. Display the response

Run the sample:

dotnet run

Advanced Usage

You can customize the agent by providing additional configuration:

using GitHub.Copilot;
using Microsoft.Agents.AI;

// Create and start a Copilot client
await using CopilotClient copilotClient = new();
await copilotClient.StartAsync();

// Create session configuration with specific model
SessionConfig sessionConfig = new()
{
    Model = "claude-opus-4.5",
    Streaming = false
};

// Create an agent with custom configuration using the extension method
AIAgent agent = copilotClient.AsAIAgent(
    sessionConfig,
    ownsClient: true,
    id: "my-copilot-agent",
    name: "My Copilot Assistant",
    description: "A helpful AI assistant powered by GitHub Copilot"
);

// Use the agent - ask it to write code for us
AgentResponse response = await agent.RunAsync("Write a small .NET 10 C# hello world single file application");
Console.WriteLine(response);

Approving or denying tool execution

The GitHub Copilot SDK owns the tool-calling loop for this provider, so approval is enforced through the SDK's native pre-execution hook rather than the Agent Framework chat-client approval round-trip.

When you register a tool wrapped in ApprovalRequiredAIFunction, GitHubCopilotAgent installs a default SessionConfig.Hooks.OnPreToolUse hook that returns "ask" for that tool and defers (null) for all other tools. The "ask" decision routes to your SessionConfig.OnPermissionRequest handler, where you approve or deny the call (this also fires even for tools configured with SkipPermission = true):

using GitHub.Copilot;

AIFunction deleteFile = AIFunctionFactory.Create(DeleteFile, "DeleteFile", "Deletes a file.");

SessionConfig sessionConfig = new()
{
    // Wrapping the tool marks it approval-required; the agent turns this into an "ask" at OnPreToolUse.
    Tools = [new ApprovalRequiredAIFunction(deleteFile)],

    // OnPermissionRequest decides the "asked" tools (and Copilot's built-in shell/file/URL prompts).
    OnPermissionRequest = (request, invocation) =>
    {
        // Surface to a human, check policy, etc.
        bool approved = AskHuman(request);
        return Task.FromResult(approved
            ? PermissionDecision.ApproveOnce()
            : PermissionDecision.Reject("Denied by user."));
    },
};

⚠️ If you provide your own OnPreToolUse hook, it takes precedence and the agent does not install its default approval hook. In that case you are fully responsible for enforcing approval — including for any ApprovalRequiredAIFunction you register (e.g. by returning a "deny" or "ask" PreToolUseHookOutput). The agent logs a warning when it detects an approval-required tool that your hook must handle.

Streaming Responses

To get streaming responses:

await foreach (AgentResponseUpdate update in agent.RunStreamingAsync("Write a C# function to calculate Fibonacci numbers"))
{
    Console.Write(update.Text);
}