copilot-ai-provider

copilot-ai-provider connects GitHub Copilot SDK to the AI SDK.

Use it in two ways:

• Standalone provider: direct Copilot client integration (Node.js/Bun runtimes).
• HTTP provider: call a hosted adapter service from any runtime.

Install

bun add copilot-ai-provider

Quick Start (Standalone)

Use this mode when your code can run a local Copilot SDK client (not browser/edge-only runtimes).

import { generateText } from "ai";
import { copilotStandalone } from "copilot-ai-provider/client/standalone";

const result = await generateText({
	model: copilotStandalone("gpt-4.1", {
		clientOptions: {
			githubToken: process.env.GITHUB_TOKEN,
		},
	}),
	prompt: "Write a haiku about TypeScript",
});

console.log(result.text);

HTTP Mode

HTTP mode is useful when you want one backend service to handle Copilot sessions and keep GitHub credentials on the server.

1) Server

Mount the provided Hono app in your framework:

import server from "copilot-ai-provider/server";

// Next.js Route Handler
export const GET = server.fetch;
export const POST = server.fetch;

Or run it as CLI server:

GITHUB_TOKEN=your-github-token API_KEY=your-api-key bunx copilot-ai-provider

Required server environment variables:

• GITHUB_TOKEN: GitHub token used by Copilot SDK.
• API_KEY: bearer token required by the HTTP adapter.

Health check endpoint:

• GET /ping → pong

2) Client

import { generateText } from "ai";
import { copilotHttp } from "copilot-ai-provider/client/http";

const result = await generateText({
	model: copilotHttp("gpt-4.1", {
		http: {
			baseUrl: "http://localhost:3000",
			apiKey: process.env.API_KEY!,
		},
	}),
	prompt: "Explain event loops in one paragraph",
});

console.log(result.text);

Exports

| Entry | Type | Exports | | --- | --- | --- | | copilot-ai-provider/client/standalone | module | copilotStandalone(modelId?, options?), createCopilotStandalone(options?), createCopilotStandaloneProvider(options?) | | copilot-ai-provider/client/http | module | copilotHttp(modelId?, options), createCopilotHttp(options), createCopilotHttpProvider(options) | | copilot-ai-provider/server | module | default Hono app instance | | copilot-ai-provider (bin) | executable | copilot-ai-provider (run with bunx copilot-ai-provider) |

Model Support

See the GitHub Copilot model catalog:

• https://docs.github.com/en/copilot/reference/ai-models/supported-models