MediaSage
github.com
- Free plan
- Windows
- macOS
- Linux
- Web
- API
- Self-hosted
What MediaSage is
MediaSage is a free, MIT-licensed, self-hosted web app that creates Plex playlists and recommends albums from a music library. Users can describe a playlist in natural language or start with a selected song, with prompts covering mood, era, instrumentation, genre, production style, and seed track. The app syncs the Plex music library to a local SQLite database for filtering and track selection. Album recommendations can use library mode for owned albums or discovery mode for albums outside the library. Filters include genres, decades, minimum rating, and excluding live recordings. Users can preview tracks, create, replace, or append Plex playlists, and send tracks to active Plex clients. MediaSage supports Google Gemini, Anthropic Claude, OpenAI GPT, Ollama, and custom OpenAI-compatible APIs, and displays estimated provider cost before generation. Setup needs Docker or Python 3.11+, a Plex server with music, a Plex token, and an AI provider key or local model. The README lists approximate capacities of 18,000 tracks for Gemini, 3,500 for Claude, and 2,300 for OpenAI. Local model support is experimental. Installation is documented for Linux, macOS, Windows, and selected NAS systems.
Who it suits
MediaSage suits Plex users who want self-hosted playlist generation and album recommendations from their own libraries. It is also aimed at people comfortable configuring Docker or Python and an AI provider.
Good to know, and what to watch
- Creates playlists from natural-language prompts
- Uses Plex library-aware recommendations and filters
- Supports multiple hosted and local AI providers
- Can preview tracks and update Plex playlists
- Runs on Linux, macOS, Windows, and listed NAS systems
- Setup requires Plex, a token, and an AI provider
- Local model support is experimental
- Provider track capacities are limited
- AI generation may have provider costs
The verdict
MediaSage combines Plex playback controls, library filtering, and AI-assisted playlist creation in a self-hosted package. Its free MIT license and local processing options appeal to technical Plex users, but setup requirements and provider limits matter.
What the maker publishes
- AI integrations
- It supports Google Gemini, Anthropic Claude, OpenAI GPT, Ollama, and custom OpenAI-compatible APIs. source
- AI integrations
- It supports Google Gemini, Anthropic Claude, OpenAI GPT, Ollama, and custom OpenAI-compatible APIs. source
- Album recommendations
- It offers recommendations in library mode for albums the user owns and discovery mode for albums they do not own. source
- Album recommendations
- It recommends albums in library mode or discovery mode and offers familiarity controls, a written pitch, and secondary picks. source
- Cost disclosure
- The app displays estimated provider cost before generation; the README lists a free Gemini API tier and describes Ollama as free. source
- Deployment
- The maker documents Docker installation and direct Python installation on Linux, macOS, and Windows, including NAS examples. source
- Filtering
- Users can filter by library genres, decades, minimum rating, and whether to exclude live recordings. source
- Filtering
- Users can filter by their library’s genre tags, decade, minimum rating, and whether to exclude live recordings. source
- Library-aware playlists
- It creates playlists from natural-language prompts or from a selected song, and says its recommendations come from tracks in the Plex library. source
- Library-aware playlists
- It offers natural-language playlist prompts and playlists based on a seed track’s mood, era, instrumentation, genre, or production style. source
- License
- The repository identifies the project as MIT licensed. source
- License
- The repository identifies the software as MIT licensed. source
- Limits
- The page lists approximate provider track capacities of 18,000 for Gemini, 3,500 for Claude, and 2,300 for OpenAI. source
- Local data
- It syncs the Plex music library to a local SQLite database for filtering and track selection. source
- Local processing
- A local SQLite cache supports library filtering and counting, and local language models are available for privacy and zero API costs. source
- Maker
- The repository owner is identified on GitHub as Eric Wilson; no headquarters or founding date is stated on the maker pages opened. source
- Plex integration
- MediaSage syncs a Plex music library and can create, replace, or append playlists and send tracks to active Plex devices. source
- Plex playback
- Users can send tracks to active Plex clients, create or update playlists, and preview tracks before saving. source
- Provider limits
- The README lists approximate maximum track counts of 18,000 for Gemini, 3,500 for Claude, and 2,300 for OpenAI, and says local model support is experimental. source
- Purpose
- MediaSage is a self-hosted web app that generates Plex playlists and album recommendations using music in the user's library. source
- Purpose
- MediaSage is a self-hosted web app that generates Plex playlists and album recommendations using music in the user’s library. source
- Requirements
- Setup requires Docker or Python 3.11+, a Plex server with music, a Plex token, and an AI provider API key or local model. source
- Requirements
- Setup requires Docker or Python 3.11+, a Plex server with music, a Plex token, and an AI provider API key or local model. source
- Security and privacy
- The maker describes local-model support as experimental and says it enables privacy and zero API costs; no security certification is stated on the page. source
Checked against the maker's pages on 9 Oct 2026.