SkillInstalls cleanly
Turn notes into audio on Spotify
Save to Spotify
Spotify's own tool: your assistant writes and narrates a briefing, lesson or meeting recap from your notes, then saves it as an episode in your Spotify.
- Works with
- Claude Code, Codex CLI
- Good for
- Daily work and research
What it does
Save to Spotify is Spotify's own command-line tool and skill. It turns things you would normally read into an episode you can listen to. You ask for a morning briefing, an article explainer, a meeting recap or a bedtime story. The assistant writes a chapter outline for you to approve, then writes the script, turns it into speech and uploads it to a show in your Spotify Library.
Spotify calls these Personal Podcasts. Spotify says they are private to you, and its help page says saved content cannot be shared. The feature is a beta for eligible Free and Premium users. Spotify's help page sets these limits:
| Account | Shows | Episodes per show |
|---|---|---|
| Free | 5 | 30 |
| Premium | 10 | 60 |
The tool does not make voices itself. You choose a voice engine:
- Kokoro: the README calls it free, local and keyless. It is a download of about 340 MB and needs Python 3.10 or newer.
- OpenAI or ElevenLabs: these use your own paid API key. The OpenAI option uses the
tts-1model, which OpenAI will remove from its API on January 6, 2027.
Things to know before you start:
- The plugin and the skills-only install do not include the program. If it is missing, the skill asks, then runs Spotify's install script (
curlintobash). The script can change your shell profile. - If you are not logged in, the skill tells the assistant to start the Spotify sign-in itself.
- Titles and descriptions cannot be changed later. To change one, delete the episode and make it again. Deleting a show deletes all its episodes.
- Your script, audio and cover image go to Spotify's servers. If you set an OpenAI key, the skill can also use it to make the cover image. Do not put passwords or confidential work in episodes.
Who it is for
People who would rather listen than read. It suits anyone who wants notes, articles or a daily plan as a private episode in their own Spotify app.
Example prompts
- "/save-to-spotify Make a 3-minute morning briefing from my calendar and to-do list."
- "Turn this article into a 10-minute explainer episode and save it to Spotify."
- "Summarise this meeting transcript as a 2-minute recap with action items."
- "Make a bedtime story about a fox called Zosia, about 5 minutes long."
How to install
curl -fsSL https://saveto.spotify.com/install.sh | bash/plugin marketplace add spotify/save-to-spotify/plugin install save-to-spotify@save-to-spotifynpx skills add spotify/save-to-spotifyHow to use it
Ask: Turn this meeting transcript into a two-minute recap with action items and save it to Spotify. You can also type /save-to-spotify in Claude Code or $save-to-spotify in Codex.
Other assistants
Codex CLI: the README says the install script makes the skill available to Codex. Codex reads skills from ~/.agents/skills, but the script links the skill there only if ~/.agents already exists.
- Install the program and the skill:
curl -fsSL https://saveto.spotify.com/install.sh | bash- Log in to Spotify:
save-to-spotify auth login- In Codex, type
$save-to-spotifyand describe the episode.
If Codex does not see the skill, install the skill on its own. The skills tool sends anonymous usage data unless you set DISABLE_TELEMETRY=1:
npx skills add spotify/save-to-spotifyGemini CLI: Spotify gives no Gemini CLI install route.
Check it works
- Check your login and voice engines:
save-to-spotify auth status
save-to-spotify tts status- In Claude Code, type
/save-to-spotifyand ask for a one-minute test episode. - The last message must say that the episode was saved as private, with a link.
The plugin is also in Anthropic's official marketplace as save-to-spotify@claude-plugins-official.
Common problems
- Login works but every call is refused —
403 ACCESS_DENIED— In issue #11 Spotify fixed this on its side. If it comes back, check that your account can use the beta, then report it on GitHub. - The episode is ready but the link does not open —
link doesn't exist— Issue #31 is still open. One user saw the episode appear about seven hours later with no change. Wait, then try the link again. - Audio does not build — the assistant reports that
ffmpegis missing — The skill needsffmpegto put the audio together. Runsave-to-spotify doctorto see what is missing. - Windows blocks the program on first run — a SmartScreen warning — The Windows build is not code-signed yet. Run the installer in Git Bash. Then right-click the file, open Properties and tick Unblock, or run
Unblock-File save-to-spotify.exein PowerShell. - You cannot rename a show — no rename option — Issue #37 is open. Show and episode details cannot be changed after you make them.
save-to-spotify@claude-plugins-officialis not found — the install fails on a new computer or in a script — Claude Code adds the official marketplace only on the first interactive session. Runclaude plugin marketplace add anthropics/claude-plugins-officialfirst.
What to watch out for
- The first line downloads and runs Spotify's install script, which puts a program on your computer; on Windows run it in Git Bash (README)
- Needs a Spotify account and a browser sign-in; the login token is saved in ~/.config/save-to-spotify/token.json
- The free local voice (Kokoro) is a download of about 340 MB (README); OpenAI or ElevenLabs voices use your own paid API key
- Spotify says episodes are private to you and its help page says saved content cannot be shared; still keep confidential work out of episodes
- Can delete your saved shows, and deleting a show also deletes all its episodes
- Your script, audio and cover image are uploaded to Spotify; OpenAI or ElevenLabs voices also send your script text to that service
- The plugin and skills routes can also run the curl install script when the program is missing; the script checks the download before it installs
- The Windows program is not code-signed yet, so SmartScreen may block the first run (README)
- The OpenAI voice option uses the tts-1 model, which OpenAI removes from its API on 2027-01-06 (cmd/tts.go, OpenAI deprecations)
- Beta limits: Free accounts get 5 shows with 30 episodes each, Premium 10 shows with 60 each (Spotify support)
- The install script offers to add a PATH line to your shell profile (install.sh)
- The install script links the skill into ~/.claude, ~/.cursor, ~/.config/opencode and ~/.agents when those folders exist
- Kokoro setup pip-installs Python packages into a new virtual environment (cmd/tts.go)
- Kokoro voice files download from a third-party GitHub repository, not from Spotify (cmd/tts.go)
- The skill makes a cover image every time; with an OpenAI key set it can use a paid image API for it (cover-image.md)
- If you are not logged in, the skill tells the assistant to start the Spotify sign-in without asking you
- Episode and show titles and descriptions cannot be changed later; you must delete and re-create (cli-usage.md)
- The token command prints your Spotify access token on screen
- The README Quick Start is a prompt that makes your assistant run the curl install script
Licence and author
- Licence
- Apache-2.0
- Author
- Spotify
- Source code
- github.com/spotify/save-to-spotify
How we tested it
We installed it on an empty test computer on 15 Oct 2026. Installed and started in a clean test system. Not yet used in a real AI session.
Not run: "curl -fsSL https://saveto.spotify.com/install.sh | bash" (not a form the sandbox runs). Not yet used on real work.
Technical details of the test
- Date
- 15 Oct 2026
- Method
- Intake sandbox run on 2026-10-15 in a clean container with no account: cloned the repository and checked its skill, agent and plugin files, ran "npx -y skills --help"; installed save-to-spotify@save-to-spotify with the Claude Code command line.
Checked again on 16 Oct 2026
Alternatives
- book-to-course: Turns long texts into a course you read in a browser, not audio.
- work-humanizer: Rewrites text so it sounds natural. It does not make audio.
FAQ
Can other people hear my episodes? Spotify says they are private to you, and saved personal content cannot be shared for now.
Do I need a paid account or API key? Spotify's help page says Free and Premium users can use it, with different limits. Kokoro needs no key. OpenAI and ElevenLabs bill you for their voices.
Does it check for updates?
Yes, in the background. Set SAVE_TO_SPOTIFY_NO_UPDATE_CHECK to turn this off.
Sources
- github.com/spotify/save-to-spotify
- raw.githubusercontent.com/spotify/save-to-spotify/main/README.md
- raw.githubusercontent.com/spotify/save-to-spotify/main/plugin/skills/save-to-spotify/SKILL.md
- raw.githubusercontent.com/spotify/save-to-spotify/main/plugin/skills/save-to-spotify/references/cli-usage.md
- raw.githubusercontent.com/spotify/save-to-spotify/main/plugin/skills/save-to-spotify/references/cover-image.md
- raw.githubusercontent.com/spotify/save-to-spotify/main/install.sh
- raw.githubusercontent.com/spotify/save-to-spotify/main/cmd/tts.go
- support.spotify.com/to/article/save-to-spotify/
- newsroom.spotify.com/2026-05-07/personal-podcasts-launch/
- developers.openai.com/api/docs/deprecations
- github.com/spotify/save-to-spotify/issues/11
- github.com/spotify/save-to-spotify/issues/31
- github.com/spotify/save-to-spotify/issues/37
- learn.chatgpt.com/docs/build-skills
- code.claude.com/docs/en/discover-plugins