Skip to content
Agnify Agnify Docs

Quickstart

Analyze a video in 3 steps: install the CLI, point it at the platform, and run one of the built-in pipelines against a local video. Total time: a few minutes plus however long your video is.

Use with your AI agent

Paste this into any AI coding agent — Claude, Cursor, Copilot, Windsurf, or similar — and it will install and configure the Agnify CLI for you automatically.

Fetch https://app.agnify.ai/docs/prompt.md

  1. Install

    Install the Agnify CLI.

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

  2. Authenticate

    agnify login opens your browser for a device-code flow and stores a refresh token in your OS keychain.

    Terminal window
    agnify login

  3. Upload + run

    One command: upload your video, kick off a pipeline against it, and block until it finishes. Exit code mirrors agnify run wait (0 completed, 1 failed, 2 timeout) so it slots into CI cleanly.

    Terminal window
    agnify ingest file ./demo.mp4 \
      --pipeline object-counter \
      --wait

  4. Read the results

    run results downloads the aggregated JSON artifact (per-frame detections + counter totals). Pipe it into jq or save with --out.

    Terminal window
    agnify run results <execution-id> \
      --out results.json
    

    # or pipe to jq:
    agnify run results <execution-id> | jq '.counter'

What you get back

Section titled “What you get back”

ingest file --wait prints the execution id, file_id, status, and viewer URL. run results gives you the full per-frame artifact; agnify events query --channel … aggregates events across runs in a metrics channel.