The serve subcommand launches a server that exposes trained models via a REST API, and that includes a web interface for exploring their predictions.

$ allennlp serve --help
usage: allennlp serve [-h] [--port PORT]

Run the web service, which provides an HTTP API as well as a web demo.

optional arguments:
-h, --help   show this help message and exit
--port PORT