Continue
Continue's VS Code and JetBrains extensions and cn CLI use a local YAML model entry to reach Rapid-MLX.
Quick start
$ npm install -g @continuedev/cli $ rapid-mlx serve qwen3.5-4b-4bit $ rapid-mlx agents continue --setup $ cn --config ~/.continue/config.yaml --auto -p "Explain this project"
Setup detects the served model, previews the YAML change, asks before writing, and checks the server. In scripts, add --yes. For another port, add --base-url http://127.0.0.1:18730/v1.
Config
Setup adds or updates a rapid-mlx model in ~/.continue/config.yaml, preserving other entries and backing up a changed file. The entry uses provider: openai, apiBase ending in /v1, and a nonempty apiKey (sk-noop for an unprotected server). rapid-mlx launch continue-dev is an alternative setup path.
If only legacy config.json exists, setup converts its settings into YAML and leaves the JSON file untouched. cn reads YAML. It can remember another selected config, so pass --config ~/.continue/config.yaml when checking this setup.
VS Code and JetBrains
Open Continue's model picker and choose rapid-mlx. Autocomplete uses a separate model role; this setup adds a chat model.
Check the connection
Ask Continue to read a local file, then verify its answer against the file. A plain chat reply establishes connectivity, while the file check establishes a tool round trip.