ipx Command Line#

ipx is the InferencePort AI command line. It is one native program that talks to everything InferencePort offers:

  • Lightning (ipx gen) - chat and generate images, video and audio with your subscription.

  • Pay-2-Go (ipx p2g) - the credit-billed API: chat, images, video, audio, 3D, embeddings, realtime voice and transcription, API keys and deployments.

  • Featherless (ipx featherless) - deploy catalog models for yourself or an organization.

  • Agents (ipx agent) - create, share and chat with managed agents.

  • Council MCP (ipx mcp council) - multi-model reasoning from the terminal.

  • AI Shield (ipx shield) - detect prompt attacks, abuse, fraud, malicious uploads, disposable emails and leaked secrets; bulk indicator checks and CI gates.

  • Organizations (ipx org) - members, teams, keys, billing, usage, audit log, SSO and policy.

  • Local models (ipx model, ipx local) - pull, manage and chat with models on your own machine, using the same Ollama engine as the desktop app.

  • Host server (ipx server) - share your local models with approved people, like the desktop app’s Host (Proxy) mode.

  • Desktop library (ipx import) - add Hugging Face Spaces and websites to the desktop app.

$ ipx auth login
$ ipx gen chat
$ ipx p2g image "a lighthouse at dawn" -o lighthouse.jpg
$ ipx auth context acme
$ ipx org members list
$ ipx model pull llama3.2 && ipx local chat llama3.2
$ ipx usage get

Designed for people and for scripts#

  • Rich, readable output in a terminal: streaming Markdown in chat, progress bars for downloads and long jobs, usage meters and tidy tables.

  • Pipe-friendly: one-shot commands write only the result to standard output; decoration goes to standard error.

  • --json on any command prints machine-readable JSON instead - see Scripting and Automation.

  • Clear errors with a hint on how to fix them (not signed in, not enough credits, organization policy, Ollama not running).

Getting started

Command guides

Reference