podpull

Download any podcast episode — Apple Podcasts, RSS, or 小宇宙/xiaoyuzhou — straight to your shell.

Pick episodes from an interactive multi-select list, with spinners and progress bars. No app, no login, no DRM. Files are named cloud-safe and (for multiple picks) grouped per show.

▶ Live animated demo & landing page: https://xiaoleiy.github.io/podpull

Demo

$ podpull get 1532755821
✓ 我們家的睡前故事 — 394 episodes
? Select episodes  (↑/↓ · space · enter)
❯ ◉ 2026-06-23  EP343 迷宮中的牛頭人
  ◯ 2026-06-16  EP342 午睡任務
  ◉ 2026-06-09  EP341 修煉龍 4
██████████████████  100%  21.2 MB  •  saved to ~/Downloads/Podcasts/我們家的睡前故事/

How it works

Apple Podcasts hosts no audio — it is a directory that points at each show's RSS feed, and every episode in that feed carries a direct <enclosure> audio URL. podpull walks that chain:

Apple show URL/id ──(iTunes Lookup API)──▶ RSS feedUrl
RSS feed          ──(<enclosure url>)────▶ direct .mp3 / .m4a
download          ──(resumable)──────────▶ <YYYY-MM-DD> - <title>.<ext>

It also resolves a pasted xiaoyuzhou episode page (via its og:audio tag) and a pasted Apple episode link (…?i=<id>, matched in the feed).

Install

# pipx (recommended) or pip — from PyPI
pipx install podpull
pip   install podpull

# or Homebrew (macOS / Linux)
brew install xiaoleiy/tap/podpull

Requires Python 3.9+. Optional: yt-dlp (deep-catalog Apple-episode fallback), ffmpeg/ffprobe (verify downloads). Also on PyPI.

Windows

Homebrew is macOS/Linux only — on Windows, install with pip/pipx (PowerShell or Windows Terminal):

# 1) Python 3.9+ (skip if you have it):
winget install Python.Python.3.12

# 2) Recommended — isolated install via pipx:
py -m pip install --user pipx
py -m pipx ensurepath          # then reopen the terminal
pipx install podpull

# …or plain pip:
py -m pip install podpull

Then run podpull in Windows Terminal or PowerShell. The interactive picker, colors, and progress bars work there; filenames are sanitized for Windows too (reserved names, illegal characters). If podpull isn't found after install, reopen the terminal (so PATH refreshes) or use py -m podpull.

Set up your AI coding agents (optional)

Teach your AI agents to use podpull so you can just ask them to grab an episode:

podpull skills install        # detected agents
podpull skills install --all  # all supported agents
podpull skills status         # see what's detected / installed

This writes podpull's instructions in each agent's native format:

Re-run after brew upgrade to refresh. Nothing is written until you run it. (Cursor has no file-based global rule — run it inside a project for a project rule, or paste the printed rule into Cursor Settings → Rules.)

Usage

podpull search "睡前故事"                 # find shows -> id, #episodes, name, author
podpull info  1532755821                  # show metadata (accepts URL, id, or RSS)
podpull list  1532755821                  # recent episodes, numbered (0 = newest)
podpull list  1532755821 --match "EP34"   # filter by title regex

Download

# Interactive picker — just give a show; pick one or many episodes with the keyboard:
podpull get 1532755821
#   ↑/↓ move · space toggle · a select-all · enter confirm

# Or select non-interactively (also used when piping / scripting):
podpull get 1532755821 --latest 1               # newest episode
podpull get 1532755821 --match "牛頭人"          # by title regex
podpull get 1532755821 --index 0,2,5            # by list number (0 = newest)
podpull get 1532755821 --latest 3 --out ~/Audio/bedtime
podpull get 1532755821 --latest 8 -j 8          # 8 episodes downloading in parallel

# Pasted single-episode links:
podpull get "https://www.xiaoyuzhoufm.com/episode/<id>"
podpull get "https://podcasts.apple.com/.../id<show>?i=<track>"

Downloads default to ~/Downloads/Podcasts (override with --out). The saved file path is printed to stdout (so you can pipe/capture it); progress and messages go to stderr. Use --no-input to never open the picker (fail instead) for scripts.

When you select several episodes they download in parallel — up to 4 at once by default. Tune it with -j/--jobs N (-j 1 forces one-at-a-time); saved paths still print to stdout in selection order.

Filenames are normalized for cloud storage — emoji and other symbols are dropped, full-width/illegal characters folded or stripped — so files upload cleanly to Google Drive, OneDrive, Dropbox, iCloud, etc. (CJK and ordinary text are kept). When you grab multiple episodes at once, they're placed in a sub-folder named after the show.

<src> accepts: an Apple show URL, a bare Apple ID, a raw RSS feed URL, an Apple episode URL (?i=), a xiaoyuzhou episode URL, or a Ximalaya album URL (ximalaya.com/album/<id>).

Podcast Index (optional)

podpull can enrich search and feed resolution with the open Podcast Index directory. Get a free API key at api.podcastindex.org and set:

export PODCASTINDEX_API_KEY=...
export PODCASTINDEX_API_SECRET=...

Without these, podpull behaves exactly as before (iTunes only).

Roadmap

• v0.1: search · info · list · download (stdlib only).
• v0.2: interactive multi-select picker, rich progress bars + spinners, colored help, scriptable stdout. Adds rich + questionary.
• v0.3: renamed podget → podpull.
• v0.4: cloud-safe filename normalization; multi-episode downloads grouped into a per-show folder.
• v0.5: pull alias for get; podpull skills install sets up integrations for Claude Code, Codex, OpenCode, and Cursor.
• v0.6 (current): robust feed parsing (RSS 2.0 / RSS 1.0 / Atom, dirty-XML recovery), verified against Chinese-market hosts, ximalaya.com/album/<id> links, optional Podcast Index (BYOK) search + feed-resolution fallback.
• next: --json output mode for scripting.
• v1+ (podpull[ai]): opt-in BYOK summarization — local transcription (faster-whisper) + your own LLM key (Anthropic/OpenAI). Fully local, private, no subscription. Cleanly isolated from the core.

Ethics & legal

podpull reads the public RSS feeds that podcasters publish for exactly this purpose, and downloads the enclosure files they distribute. Respect each show's copyright and terms — download only what you're entitled to, for personal use.

License

MIT © xiaoleiyu