Skip to main content

VS Code and GitHub Copilot with Verdictan

Port 41002 is an LLM API endpoint, not a forward HTTP proxy. Do not put it in http.proxy, HTTP_PROXY, or HTTPS_PROXY. Those settings cause Copilot to use CONNECT/forward-proxy behavior that the Verdictan gateway does not implement.

VS Code offers two supported Verdictan connection paths.

Add Verdictan MCP tools​

Create .vscode/mcp.json. VS Code uses servers, not Cursor's mcpServers shape. It can store the token through a password input:

{
  "inputs": [
    {
      "type": "promptString",
      "id": "verdictan-token",
      "description": "Verdictan MCP client token",
      "password": true
    }
  ],
  "servers": {
    "verdictan": {
      "type": "http",
      "url": "https://<published-hostname>/mcp",
      "headers": {
        "Authorization": "Bearer ${input:verdictan-token}"
      }
    }
  }
}

Run MCP: List Servers to start the server and examine its output. See the VS Code MCP guide for trust and organization controls.

Route VS Code chat through a custom endpoint​

VS Code's Custom Endpoint model provider can show Chat Completions, Responses, or Messages API options. All three Verdictan request families use the documented access, input, tool, output, and audit stages. Select the API type that the editor and configured provider support.

Open Chat: Manage Language Models. Select Add Models → Custom Endpoint. Configure the Verdictan URL, client token, model ID, API type, and capabilities.

For an OpenAI-compatible model, use the gateway route that the editor must use in the editor's configuration UI, such as:

http://127.0.0.1:41002/v1/chat/completions

VS Code's BYOK custom models apply to chat and utility tasks. GitHub documents that semantic search, inline suggestions, and some embedding-backed features continue to use GitHub services. See the VS Code language-model guide for this boundary.

What is not governed​

MCP does not reroute Copilot model traffic that GitHub hosts. A custom chat model does not mean that inline suggestions use it. Verdictan can govern only requests that target its endpoint.

After you select the custom model, verify with verdictan events tail --since 10m --follow.

Next steps​