Add OpenAI-compatible HTTP server mode

[dhaern]

What feature would you like to see?

Hi! I’m using code on Windows and I’m trying to connect it to Open WebUI (or any OpenAI-compatible UI). Right now I can run code app-server, but I don’t see it listening on any TCP port (verified with Get-NetTCPConnection), and I can’t find documented flags to bind a host/port or expose an HTTP API.

Could you add a simple HTTP server mode that exposes an OpenAI-compatible API, so code can be used directly as a backend (no custom wrappers)?

Request:

• Something like: code serve --host 127.0.0.1 --port 8001 (or app-server with --host/--port)

• Endpoints:
  GET /v1/models
  POST /v1/chat/completions (support stream: true SSE, and stream: false)

• Optional/required Bearer auth (API key), Session handling-Respect an incoming session_id (or generate one) to preserve CLI/session context for agentic workflows

This would make integration with "chat" UIs like Open WebUI and similar tools straightforward while keeping all the “agentic” CLI functionality under the hood.