Inference CLI | Roboflow Docs

Command-line tool for running the Inference Server, processing data with Workflows, benchmarking, making predictions, and deploying to the cloud.

The Roboflow Inference CLI is a command-line interface for the inference ecosystem. Use it to:

run and manage the Inference Server locally
process data with Workflows
benchmark inference performance
make predictions from your models
deploy the Inference Server to the cloud

Installation

pip install inference-cli

If you have installed the inference Python package, the CLI is already included.

This is the Inference CLI (inference ...), which manages the open source Inference Server. It is a different tool from the Roboflow CLI (roboflow ...), which manages platform resources such as projects, datasets, and versions.

Supported devices

The Inference CLI currently supports the following device targets:

x86 CPU
ARM64 CPU
NVIDIA GPU (including Jetson)

For Jetson-specific Inference Server images, see the Inference Server page or pull the images directly.

Commands

Command	Description
`inference server`	Start, stop, and check the local Inference Server.
`inference infer`	Run predictions on images, directories, and video files.
`inference workflows`	Process images and videos with a Workflow.
`inference benchmark`	Benchmark the Inference Server or the `inference` Python package.
`inference cloud`	Deploy the Inference Server to AWS or GCP virtual machines.