agnify

Agnify CLI

agnify is the command-line client for Agnify, a video-analysis platform. Drive pipelines, upload media, run executions, manage cameras, and query events — all from your terminal.

Install

One line (installs uv if needed, then the CLI):

curl -LsSf https://app.agnify.ai/install.sh | sh

Or, if you already have a Python toolchain:

uv tool install 'agnify[keyring]'   # recommended — OS keychain for tokens
# or
pipx install 'agnify[keyring]'
# or
pip install 'agnify[keyring]'

The [keyring] extra stores your refresh token in the OS keychain (macOS Keychain, Secret Service, Windows Credential Manager). Plain agnify also works — it falls back to a 0600 file under ~/.config/agnify/.

Authenticate

agnify login        # device-code flow in your browser
agnify whoami       # confirm the active user + org

For CI / cron / integrations, skip login and set an org-scoped API key:

export AGNIFY_API_KEY=ak_…          # minted in the Agnify dashboard
export AGNIFY_API_BASE=https://app.agnify.ai

Self-hosted? Point the CLI at your backend with AGNIFY_API_BASE.

Use

agnify nodes list                   # the pipeline node menu
agnify ingest file ./clip.mp4 --pipeline <template> --wait    # upload + run + wait for results
agnify run ls                       # list executions
agnify events tail                  # stream live pipeline events

Full command reference: https://app.agnify.ai/docs/cli

To import historical footage into a camera, supply its original start time explicitly, including Z or a UTC offset:

agnify camera stream cam-EXAMPLE --source ./footage.mp4 \
  --recorded-at 2026-08-01T09:00:00Z

Each segment is timestamped at that start time plus its actual position in the video. This option requires a local video file; it does not read recording dates from video metadata. Without it, streaming uses the current wall clock.

License

Apache-2.0 © 2026 Fewbits Inc. (Agnify).