Spoken — Claude Code MCP server: podcast transcripts
Hands your agent Markdown transcripts of published podcasts, with the actual names of the people talking. Useful when you need an episode's text and would rather skip the audio.
claude mcp add --transport http spoken https://spoken.md/mcpWhat Spoken does
Spoken is an API for pulling transcripts of podcasts that are already out. Nothing gets uploaded, no diarization runs on your side: the service fills in real speaker names in place of anonymous labels, adds timestamps, and returns Markdown shaped for LLM context and RAG chunking.
The local build ships as spoken-mcp and runs through npx; the hosted one lives at https://spoken.md/mcp. Either way you get eight tools: search episodes by text or a pasted Spotify/YouTube link, list a show's catalog, fetch a transcript, check credits, follow shows, and see new episodes. A key goes in SPOKEN_API_KEY. Leave it out and the demo key pt_demo takes over — search and episode listing stay fully usable, while transcripts stop at the demo episode.
Credits are pay-per-use, no subscription attached. Fetching the same transcript again costs nothing, and failed calls are not billed. Keeping a knowledge base fresh comes down to calling list_new_episodes on a schedule and then get_transcript on whatever it returns.
What it helps with
- get an episode's text without uploading audio
- replace anonymous speaker labels with real names
- split a transcript into chunks for a vector store
- track new episodes of shows you follow
- find an episode from a Spotify or YouTube link
Example prompts for the agent
- Find a podcast episode about sleep and show me its transcript
- Pull the Markdown transcript for this episode with speaker names
- List new episodes of my followed shows that I haven't fetched yet
How to install Spoken
1Platform CLI: remote server
claude mcp add --transport http spoken https://spoken.md/mcp2Platform CLI
claude mcp add --env SPOKEN_API_KEY=<SPOKEN_API_KEY> --transport stdio spoken -- npx -y spoken-mcpReplace <…> with your own values
1Platform CLI: remote server
codex mcp add spoken --url https://spoken.md/mcp2Platform CLI
codex mcp add spoken --env SPOKEN_API_KEY=<SPOKEN_API_KEY> -- npx -y spoken-mcpReplace <…> with your own values
Config file
{
"mcpServers": {
"spoken": {
"url": "https://spoken.md/mcp"
}
}
}{
"mcpServers": {
"spoken": {
"command": "npx",
"args": [
"-y",
"spoken-mcp"
],
"env": {
"SPOKEN_API_KEY": "<SPOKEN_API_KEY>"
}
}
}
}Replace <…> with your own values
Config file
{
"servers": {
"spoken": {
"type": "http",
"url": "https://spoken.md/mcp"
}
}
}{
"servers": {
"spoken": {
"type": "stdio",
"command": "npx",
"args": [
"-y",
"spoken-mcp"
],
"env": {
"SPOKEN_API_KEY": "<SPOKEN_API_KEY>"
}
}
}
}Replace <…> with your own values
1Platform CLI: remote server
gemini mcp add -t http spoken https://spoken.md/mcp2Platform CLI
gemini mcp add -e SPOKEN_API_KEY=<SPOKEN_API_KEY> spoken npx -y spoken-mcpReplace <…> with your own values
Step by step, with commands for each platform: How to add an MCP server to Claude Code, Codex, Cursor, VS Code, Gemini CLI and more
Questions about Spoken
How do I install Spoken in Claude Code?
For the hosted server run claude mcp add --transport http spoken https://spoken.md/mcp. For the local one run claude mcp add --env SPOKEN_API_KEY=<SPOKEN_API_KEY> --transport stdio spoken -- npx -y spoken-mcp.
What does Spoken do?
It returns transcripts of published podcasts as Markdown with real speaker names and timestamps. It also searches episodes, tracks which shows a key follows, and lists new episodes.
Do I need an API key?
Yes, it goes in SPOKEN_API_KEY. Without one the demo key pt_demo is used: search and episode listing work, but transcripts are limited to the demo episode.
