generated: '2026-06-20'
method: searched
source: https://docs.ngc.nvidia.com/cli/index.html; https://docs.api.nvidia.com/nim/docs/deployment
notes: >-
  NIM has no single "nim" binary; the first-party CLI for pulling, listing, and running NIM
  container microservices is the NGC CLI (ngc), which manages images/models in the NVIDIA GPU
  Cloud registry (nvcr.io). Actual inference is then driven over the OpenAI-compatible REST/gRPC
  surface. The docker CLI runs the pulled NIM image. The binary itself is not on PyPI/npm; it
  is distributed from NGC.
cli:
  name: ngc
  binary: ngc
  distribution: https://org.ngc.nvidia.com/setup/installers/cli
  auth:
    method: NGC Personal API Key
    config: ngc config set
    registry_login: 'docker login nvcr.io -u $oauthtoken -p $NGC_API_KEY'
  command_groups:
    - group: config
      purpose: Configure API key, org/team, and output format.
      examples:
        - ngc config set
    - group: registry image
      purpose: Discover and inspect NIM container images.
      examples:
        - 'ngc registry image list --format_type csv nvcr.io/nim/meta/*'
        - ngc registry image info nvcr.io/nim/nvidia/fourcastnet:1.1.0
    - group: registry model
      purpose: List and download model artifacts backing a NIM.
      examples:
        - ngc registry model list
  typical_flow:
    - ngc config set                              # store API key + org
    - docker login nvcr.io -u '$oauthtoken' -p $NGC_API_KEY
    - ngc registry image list 'nvcr.io/nim/meta/*'
    - docker run --gpus all -e NGC_API_KEY -p 8000:8000 nvcr.io/nim/meta/llama-3.1-70b-instruct:latest