A local interface for GitHub Copilot built on the official VS Code Language Models API.
Copilot Bridge lets you access your personal Copilot session locally through an OpenAI-compatible interface — without calling any private GitHub endpoints. It’s designed for developers experimenting with AI agents, CLI tools, and custom integrations inside their own editor environment.
API Surface: Uses only the public VS Code Language Model API (
vscode.lm) for model discovery and chat. No private Copilot endpoints, tokens, or protocol emulation.
- Local HTTP server locked to
127.0.0.1 - OpenAI-style
/v1/chat/completions,/v1/models, and/healthendpoints - SSE streaming for incremental responses
- Real-time model discovery via VS Code Language Model API
- Concurrency and rate limits to keep VS Code responsive
- Mandatory bearer token authentication with
HTTP 401 Unauthorizedprotection - Lightweight Polka-based server integrated directly with the VS Code runtime
- Uses only the public VS Code Language Models API.
- Does not contact or emulate private GitHub Copilot endpoints.
- Requires an active GitHub Copilot subscription.
- Subject to GitHub Terms of Service and the Github Acceptable Use Policy.
- Intended for personal, local experimentation only.
- No affiliation with GitHub or Microsoft.
❗ The author provides this project as a technical demonstration. Use responsibly and ensure your own compliance with applicable terms.
| ✅ Supported | 🚫 Not Supported |
|---|---|
| Local, single-user loopback use | Multi-user or shared deployments |
| Testing local agents or CLI integrations | Continuous automation or CI/CD use |
| Educational / experimental use | Public or commercial API hosting |
Copilot Bridge was built to demonstrate how VS Code’s Language Model API can power local-first AI tooling.
It enables developers to reuse OpenAI-compatible SDKs and workflows while keeping all traffic on-device.
This is not a Copilot proxy, wrapper, or reverse-engineered client — it’s a bridge built entirely on the editor’s public extension surface.
This software is provided as is for research and educational purposes.
Use at your own risk.
You are solely responsible for ensuring compliance with your Copilot license and applicable terms.
The author collects no data and has no access to user prompts or completions.
- Visual Studio Code Desktop with GitHub Copilot signed in
- (Optional) Node.js 18+ and npm for local builds
- Install from the Visual Studio Marketplace or load the
.vsix. - Set Copilot Bridge › Token to a secret value (Settings UI or JSON). Requests without this token receive
401 Unauthorized. - Open the Command Palette → “Copilot Bridge: Enable” to start the bridge.
- Check status anytime with “Copilot Bridge: Status” or by hovering the status bar item (it links directly to the token setting when missing).
- Keep VS Code open — the bridge runs only while the editor is active.
Replace PORT with the one shown in “Copilot Bridge: Status”. Use the same token value you configured in VS Code:
export PORT=12345 # Replace with the port from the status command
export BRIDGE_TOKEN="<your-copilot-bridge-token>"List models:
curl -H "Authorization: Bearer $BRIDGE_TOKEN" \
http://127.0.0.1:$PORT/v1/modelsStream a completion:
curl -N \
-H "Authorization: Bearer $BRIDGE_TOKEN" \
-H "Content-Type: application/json" \
-d '{"model":"gpt-4o-copilot","messages":[{"role":"user","content":"hello"}]}' \
http://127.0.0.1:$PORT/v1/chat/completionsUse with OpenAI SDK:
import OpenAI from "openai";
if (!process.env.BRIDGE_TOKEN) {
throw new Error("Set BRIDGE_TOKEN to the same token configured in VS Code settings (bridge.token).");
}
const client = new OpenAI({
baseURL: `http://127.0.0.1:${process.env.PORT}/v1`,
apiKey: process.env.BRIDGE_TOKEN,
});
const rsp = await client.chat.completions.create({
model: "gpt-4o-copilot",
messages: [{ role: "user", content: "hello" }],
});
console.log(rsp.choices[0].message?.content);The extension uses VS Code’s built-in Language Model API to select available Copilot chat models.
Requests are normalized and sent through VS Code itself, never directly to GitHub Copilot servers.
Responses stream back via SSE with concurrency controls for editor stability.
import * as vscode from "vscode";
const models = await vscode.lm.selectChatModels({
where: { vendor: "copilot", supports: { reasoning: true } }
});
const model = models[0] ?? (await vscode.lm.selectChatModels({}))[0];
if (!model) throw new Error("No language models available (vscode.lm)");
const stream = await model.sendRequest(
{ kind: "chat", messages: [{ role: "user", content: "hello" }] },
{ temperature: 0.2 }
);
// Stream chunks → SSE to localhost client; no private Copilot protocol used.| Setting | Default | Description |
|---|---|---|
bridge.enabled |
false | Start automatically with VS Code |
bridge.port |
0 | Ephemeral port |
bridge.token |
"" | Bearer token required for every request (leave empty to block API access) |
bridge.historyWindow |
3 | Retained conversation turns |
bridge.maxConcurrent |
1 | Max concurrent requests |
bridge.verbose |
false | Enable verbose logging |
ℹ️ The bridge always binds to
127.0.0.1and cannot be exposed to other interfaces.
💡 Hover the status bar item to confirm the token status; missing tokens show a warning link that opens the relevant setting.
- Enable
bridge.verbose. - Open View → Output → “Copilot Bridge”.
- Observe connection events, health checks, and streaming traces.
⚠️ This extension is intended for localhost use only.
Never expose the endpoint to external networks.
- Loopback-only binding (non-configurable)
- Mandatory bearer token gating (requests rejected without the correct header)
- Telemetry: none collected or transmitted.
- v1.2.0 – Authentication token now mandatory; status bar hover warns when missing
- v1.1.1 – Locked the HTTP server to localhost for improved safety
- v1.1.0 – Performance improvements (~30%)
- v1.0.0 – Modular core, OpenAI typings, tool-calling support
- v0.2.2 – Polka integration, improved model family selection
- v0.1.0–0.1.5 – Initial releases and bug fixes
Pull requests and discussions are welcome.
Please open an issue to report bugs or suggest features.
Apache 2.0 © 2025 [Lars Baunwall]
Independent project — not affiliated with GitHub or Microsoft.
For compliance or takedown inquiries, please open a GitHub issue.
No. Copilot Bridge is designed for localhost-only, single-user, interactive use.
Running it on a shared host or exposing it over a network would violate its intended scope and could breach the Copilot terms.
The host is bound to 127.0.0.1 (non-configurable).
No. The bridge never transmits telemetry, prompts, or responses to any external service.
All traffic stays on your machine and flows through VS Code’s built-in model interface.
The /health endpoint will report a diagnostic reason such as copilot_unavailable or missing_language_model_api.
This means VS Code currently has no accessible models via vscode.lm. Once Copilot becomes available again, the bridge will resume automatically.
Yes, if other providers register with vscode.lm. The bridge will detect any available chat-capable models and use the first suitable one it finds.
Reverse-engineered proxies call private endpoints directly or reuse extracted tokens.
Copilot Bridge does neither—it communicates only through VS Code’s sanctioned Language Model API, keeping usage transparent and compliant.