Skip to content

Copilot Chat Extension Model

Key Points

  • Copilot extensions add custom commands / context to GitHub Copilot Chat in VS, VS Code, JetBrains.
  • Two extension types: chat participants (@myparticipant) and slash commands (/mycommand).
  • Build for IDE-specific or cross-IDE.
  • Use cases: project-specific assistance, internal docs lookup, custom code generation.

Concept

User: @myproject /createcontroller User

Copilot Chat → @myproject extension → produces code → user accepts.

Extensions register with the IDE; user invokes via @ or /.

VS Code Copilot extension

// package.json
{
  "name": "myproject-copilot",
  "main": "./extension.js",
  "contributes": {
    "chatParticipants": [{
      "id": "myproject",
      "name": "myproject",
      "fullName": "MyProject Assistant",
      "description": "Helps with MyProject conventions",
      "commands": [
        { "name": "createcontroller", "description": "Generate a controller" }
      ]
    }]
  }
}

// extension.js
const vscode = require('vscode');

function activate(ctx) {
    const handler = async (request, context, stream, token) => {
        if (request.command === 'createcontroller') {
            stream.markdown('Generating controller for: ' + request.prompt);
            // call Copilot API or your own
        }
    };

    const participant = vscode.chat.createChatParticipant('myproject', handler);
    participant.iconPath = vscode.Uri.joinPath(ctx.extensionUri, 'icon.png');
}

module.exports = { activate };

Visual Studio Copilot extension

For Visual Studio (full IDE): use Visual Studio extension SDK + Copilot APIs (preview/GA varies).

C# / VSIX based.

JetBrains Copilot

JetBrains has its own AI / Copilot. Extension model differs.

What can extensions do

  • Read open files / selection.
  • Stream Markdown / code to chat.
  • Reference files via references.
  • Call external APIs (your backend).
  • Use Copilot's LLM (passes through).

Example: project conventions

const handler = async (request, context, stream) => {
    if (request.command === 'review') {
        const editor = vscode.window.activeTextEditor;
        const code = editor?.document.getText() ?? '';

        // Send to your backend with project context
        const response = await fetch('https://my-backend/review', {
            method: 'POST',
            body: JSON.stringify({ code, conventions: ourConventions })
        });

        stream.markdown(await response.text());
    }
};

Use cases

  • Internal SDK docs: query company knowledge base.
  • Project-specific generators: generate code matching your conventions.
  • Architectural review: against ADRs, design rules.
  • PR templates: generate description.

Distribution

  • Marketplace: VS Code marketplace, JetBrains marketplace.
  • Internal: company VSIX feed; private extensions.

.NET-specific examples

  • Aspire scaffolder: generate AppHost integrations.
  • Entity generator: from DB schema.
  • Test scaffolder: from Razor views.

Senior considerations

  • Privacy: extension can access source code; handle sensitive data carefully.
  • Auth: tokens for backend calls.
  • Testing: hard; mock chat API.
  • Versioning: extensions break with IDE updates.

Cross-references