BokeBox MCP
Self-hosted podcast studio whose built-in MCP endpoint creates episodes from links and text and manages schedules.
- 104 ★ on the repository
- Last commit
- LGPL-3.0
Extended access
Why this level
- The MCP token can create and delete episodes and change schedules.
- Texts and links go to the configured AI, recognition and voice services.
Install
- Works in
Manual install
{"mcpServers":{"bokebox":{"url":"http://localhost:8787/mcp","headers":{"Authorization":"Bearer <token from Settings>"}}}}The README Cursor configuration block. Start the server first with `./start.sh` or Docker.
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- The MCP token can create and delete episodes and change schedules.
- Texts and links go to the configured AI, recognition and voice services.
This is third-party code. Review the repository files before installing.
What it does
BokeBox turns links, videos, articles, notes and meeting recordings into private podcasts: speech recognition, rewriting into a spoken script with a host persona, voice synthesis, a cover and review flashcards. It all runs on your own server with SQLite and local storage. The built-in `POST /mcp` endpoint with a long-lived token from settings gives an agent the tools `create_podcast_from_url`, `create_podcast_from_text`, `list_jobs`, `get_job`, `retry_job`, `list_library` and schedule management. Plugins extend sources, speech recognition, voice synthesis and schedules.
Who it is for. Self-hosters who listen to processed material instead of reading it and manage episodes from Cursor, Claude Desktop or Codex.
Good fit when
- A long article by link must become an episode to listen to on the way.
- After a meeting an agent sends the transcript to voice synthesis with `create_podcast_from_text`.
- Source subscriptions run on a schedule and job status is checked through `list_jobs`.
- A personal podcast with a custom voice and host persona is needed without a cloud platform.
Not a fit when
- A public podcast platform or multi-user service is needed; the project targets a single user.
- There is no wish to run a server and connect AI service keys.
- Material must not be passed to external speech recognition and voice services.
Example request
Create a podcast from this link, then show the job status and the link to the finished episode.MCP
- Transport
- http
- Authentication
- API key
Environment variables
PUBLIC_BASE_URL- Public server address for install links behind a reverse proxy
Limitations
A server of your own or Docker and AI service keys in `.env` are required. The project targets a single user with SQLite. MCP runs over HTTP with a Bearer token, so do not expose the port to the internet without a reverse proxy. The authors stipulate that enabling plugins, MCP and schedules is at your discretion, with responsibility for data and account bans resting on you. License: LGPL-3.0. Keys for the recognition, AI and voice services are your own, and without them episodes are not created.
How to disable. Remove the `bokebox` block from the client configuration and stop the server or Docker container. Delete the SQLite database and media files separately.
FAQ
How do I connect Cursor?
Add a server with a `url` like `http://localhost:8787/mcp` and an `Authorization: Bearer <token>` header. The token is in the MCP section of settings.
Which tools does the server expose?
Creation from a link or text, `list_jobs`, `get_job`, `update_job`, `retry_job`, `delete_job`, `list_library`, schedule management and `get_system_health`.
How do I start the server?
Run `./start.sh` for development on port 5173 or `./start.sh docker` with the `ghcr.io/vastsa/bokebox:latest` image on port 8787.
How do I set a public address behind a proxy?
The `PUBLIC_BASE_URL` variable sets the address for install links behind a reverse proxy.
Can sources be extended?
Yes. Plugins for sources, speech recognition, voice synthesis and schedules are enabled in settings by rescanning or uploading a zip.
Discussion
No questions or reviews yet.
Related
MCP serversA local sound studio with an MCP server: an agent makes effects, music, songs and speech on open-weight models
A Podwise console client that gives an agent transcripts, summaries, chapters and answers for podcast episodes and YouTube videos
The audio package runs an MCP server for sound work: loudness, spec checks, denoise, EQ, cuts and tempo.
MCP server for Audacity 3.x with 144 tools and 9 pipelines for audio cleanup, mastering and transcription, all processed locally.
A command-line tool that finds podcast episodes on Apple Podcasts, RSS or Xiaoyuzhou and saves them as files, with agent support.
A text-to-speech MCP server that turns text and multi-voice scripts into audio through ElevenLabs and keeps a job history.
Comparison with similar tools
Cards that share the main search query, and how they differ.
| Tool | Official | Authentication | Repository stars | Main difference |
|---|---|---|---|---|
| BokeBox (this card) | no | API key | 104 | |
| Podcast to Course | no | 115 | a neighboring podcast tool; BokeBox runs on your server and is driven by an agent through MCP | |
| Video Podcast Maker | no | 1,720 | builds video podcasts; BokeBox makes spoken episodes from links and text and runs schedules |
README in short
The project starts with one launch command or a Docker image. Besides MCP there is a web interface with a player, albums and review flashcards.