Agent guide · rapid-mlx 0.15.7 · ← Back to README

Cursor

Cursor is a closed-source IDE that supports a custom OpenAI base URL for its chat/composer surface. Cursor calls that URL from its own servers, so rapid-mlx has to be reachable over public HTTPS, behind an API key.

Quick start

Needs: Cursor, a public HTTPS tunnel to your Mac's port 8000, and an API key. Cursor sends custom-model requests from its own servers, so it cannot reach localhost.

$ export RAPID_MLX_API_KEY="$(openssl rand -hex 24)"
$ rapid-mlx serve qwen3.6-35b-4bit             # uses RAPID_MLX_API_KEY
# expose port 8000 over HTTPS with a tunnel, then:
$ rapid-mlx launch cursor --server-url https://<your-tunnel-host>

Then pick the model in Cursor's model dropdown and send a chat message. A reply means it works.

TIER-2 AGENT   Wire: /v1/chat/completions (IDE chat/composer path) · Setup: rapid-mlx launch cursor --server-url https://… · Matrix cell: no dedicated cell (Cursor's agent path is locked to Cursor's backend — see gotcha below).
Tier-2 because only the IDE-chat surface is integrable and it needs the GUI app — the CLI/agent path can't point at a custom endpoint, so it can't get the per-release headless guarantee a Tier-1 agent gets.

Install

Download from cursor.com (closed-source vendor binary).

Config

Cursor routes custom-model (BYOK) requests through its own backend, so the URL you give it must be reachable from the internet over HTTPS. A tunnel that forwards to port 8000 on your Mac works. Never expose a server without a key: start it with one, and rapid-mlx launch cursor refuses to run without RAPID_MLX_API_KEY.

$ export RAPID_MLX_API_KEY="$(openssl rand -hex 24)"
$ rapid-mlx serve qwen3.6-35b-4bit          # serve reads RAPID_MLX_API_KEY
$ rapid-mlx launch cursor --server-url https://<your-tunnel-host> --dry-run
$ rapid-mlx launch cursor --server-url https://<your-tunnel-host>

launch cursor writes Cursor's user settings.json (~/Library/Application Support/Cursor/User/), appending /v1 to the URL and keeping your other settings. It rejects localhost, private addresses and plain HTTP before writing anything.

Run

With the server and tunnel up, open Cursor and pick the model from the model dropdown. Chat and composer requests then go through the tunnel to your Mac.

Gotchas

See also