yt-research

Export and analyze a YouTube channel's complete public video history from your terminal.

Explore channel metadata, find top videos, and export clean CSV or JSON for spreadsheets, scripts, and agents.

Quick start · Commands · Documentation · Contributing

[!IMPORTANT] yt-research is under active development. The 0.1.x interface may evolve before 1.0.

✨ Why yt-research?

The YouTube website is built for watching. yt-research is built for answering questions about a channel's public catalog.

discover  →  collect  →  filter  →  rank  →  export

It uses the official YouTube Data API v3 with your own API key and requests only public data. Ambiguous channel searches return candidates instead of silently choosing one.

⚡ Quick start

1 · Install

Use pipx with Python 3.11+ on macOS or Linux:

pipx install yt-research

2 · Authenticate

Add your YouTube Data API key to your operating system's native secret store:

yt-research auth set
yt-research auth status

3 · Research

Replace the placeholder with a public channel handle:

# See the newest uploads
yt-research videos latest @examplecreator --limit 10

# Find the most-viewed videos published in 2026
yt-research videos top @examplecreator --year 2026

# Search titles and save the results
yt-research videos list @examplecreator --match "tutorial" --format csv --output videos.csv

# Summarize the latest 12 calendar months and find breakout uploads
yt-research channel analyze @examplecreator --limit 10

Terminal output defaults to a readable table. Redirected output automatically switches to JSON, while diagnostics stay on stderr:

yt-research videos list @examplecreator > videos.json

[!TIP] Need an API key? Follow the step-by-step API key setup guide. In CI or headless Linux, set YT_RESEARCH_API_KEY instead; it takes precedence over the native secret store.

uv tool install yt-research

uvx yt-research --help

git clone https://github.com/vedntp/yt-research.git
cd yt-research
uv sync
uv run yt-research --help

🧭 Commands at a glance

Video commands support --match, --year, --from, --to, --sort, --limit, --format, and --output. Run any command with --help or see the full command reference.

channel analyze defaults to the latest 12 calendar months. Use --months N, --year YYYY, --from/--to, or --all to choose a different window. Its aggregate summary always includes every matching upload in that window; --limit only limits the breakout-video table.

Research recipes

yt-research videos list @examplecreator --match "telescope" --sort published-asc

yt-research videos top @examplecreator --year 2026 --limit 25

yt-research channel analyze @examplecreator --months 12 --match "tutorial"

yt-research videos list @examplecreator --from 2025-01-01 --format json --output videos.json

🔌 Built for people and programs

Interactive results are formatted as a styled terminal table:

_{Illustrative demo; titles and statistics are not actual MKBHD data · open full size}

The same commands fit cleanly into pipelines and spreadsheet workflows:

# Pipeline → versioned JSON
yt-research videos list @examplecreator --year 2026 | jq '.items[].title'

# Spreadsheet → fixed-column CSV
yt-research videos top @examplecreator --limit 50 --format csv --output top-videos.csv

JSON output includes a schema version, command metadata, the resolved channel, the effective query, result items, and request counts. Analysis reports add typed summary and publication-cohort sections while retaining the same envelope. See the JSON output contract for the stable shape.

📚 Documentation

Scope and privacy

yt-research does not use OAuth or access private account data. It does not download videos, retrieve transcripts or comments, classify Shorts, or run as a hosted service. API calls consume quota from the Google Cloud project associated with your key. Analysis reports may traverse every upload in the selected window because aggregate metrics must not be truncated; use a bounded date window and title match to keep quota use focused.

Contributing

Contributions are welcome. Read the contributing guide, Code of Conduct, and Security Policy before opening a contribution.

License

Released under the MIT License.

_{Built for curious people, repeatable research, and clean data.}