io.github.spokenmd由 GitHub 帳號 spokenmd 發布v0.4.3更新於 Oct 24, 2026
Podcast transcripts as clean Markdown with real speaker names — via the Spoken API.
Spoken 讓助理搜尋播客，並取得帶有真實講者姓名的乾淨 Markdown 逐字稿，還能追蹤節目、找出新單集。
Web executable，透過 Streamable HTTP。 遠端服務在工作區中設定後即可從網頁執行環境執行。
把這段提示詞貼上到 Claude Code、Codex、Cursor 等代理程式，或 SourceWeft 對話中。代理程式會讀取該伺服器的安裝指南，先向你展示發布者、端點和登入方式，經你同意後再把它新增到你的客戶端。金鑰不會寫進任何命令。
閱讀 https://sourceweft.com/mcp/io.github.spokenmd%2Fspoken/install.md，並按其中說明把這個 MCP 伺服器接入我的客戶端。修改任何設定之前，先告訴我它的發布者、端點和登入方式，等我確認。不要把 API 金鑰寫進命令或我會提交的檔案；問我金鑰應從哪裡取得。把它新增到你客戶端的 mcpServers 設定中。
{
"mcpServers": {
"spoken": {
"type": "http",
"url": "https://spoken.md/mcp"
}
}
}Spoken is a transcript API that turns any published podcast into clean Markdown with real speaker names — not "Speaker 1." One API call returns named, timestamped text, ready for LLMs, RAG pipelines, summarizers, and search. This repo also ships spoken-mcp, an MCP server that gives Claude Desktop, Claude Code, Cursor and Cline the same transcripts as tools — see Use as an MCP server.
It's a transcript retrieval API, not a speech-to-text service: it works on already-published podcasts, so you skip uploading audio, running diarization, and mapping anonymous speaker labels by hand. For published shows that's typically 5–10× cheaper than running the audio through a transcription service.
agents.md, llms.txt, and an OpenAPI specGet a key at spoken.md — or try it free with the demo key pt_demo (search works fully; transcripts limited to the demo episode).
The transcript comes back as Markdown with named speakers and timestamps:
Auth is the x-api-key header. Responses include X-Credits-Remaining and X-Credits-Charged. See agents.md for the full error table and response shapes.
examples/podcast_summarizer.py — fetch a transcript and summarize itexamples/rag_pipeline.py — chunk a transcript for a vector store / RAGexamples/quickstart.sh — search → transcript in two curl callsexamples/archive-show.sh — archive a show's entire back catalogue, one file per episodeThe spoken-md package wraps the API with no dependencies outside the standard library, and installs a spoken-md command.
archive(podcast_id, skip=...) walks a whole show and is resumable; errors are typed by status (PaymentRequired carries the top-up URL, NotFound means no transcript). See python/README.md.
Spoken is a Model Context Protocol server two ways: hosted at https://spoken.md/mcp, and as spoken-mcp, the package in this repo, for clients that run a server locally (Claude Desktop, Cursor, Cline, …). Both provide the same eight tools:
Keeping a knowledge base current is list_new_episodes on a schedule and get_transcript on what it lists: each fetch raises that show's floor.
Nothing to install. Point a client that connects to a URL at https://spoken.md/mcp (Streamable HTTP):
The client asks you to sign in the first time a tool needs your key. To skip that, send the key as a header yourself: Authorization: Bearer pt_your_key.
A client that can only sign in with OAuth, such as ChatGPT in developer mode, is sent to a page where you paste your key once. Steps for each app: spoken.md/mcp-server. With no key at all, searching, listing a show and the demo episode work.
Add it to your MCP client config (e.g. Claude Desktop's claude_desktop_config.json):
SPOKEN_API_KEY defaults to pt_demo (search works fully; transcripts limited to the demo episode). Get a real key at spoken.md.
Run from source instead:
Spoken is designed to be called by agents. Point your agent at the Agent Skill (also served at https://spoken.md/.well-known/skills/spoken-md/SKILL.md), or hand it agents.md. The OpenAPI spec makes it easy to wrap as a tool for any function-calling or MCP-compatible client (Claude, GPT, Cursor).
Pay-per-use credits, no subscription. New keys: 100 for $15, 500 for $50, 2,000 for $160. Machine-readable at spoken.md/pricing.md.
Spoken is built and maintained at spoken.md.
來源：README.md，提交 079853d
Search podcasts
Search published podcast episodes by text query, or paste an episode URL (Spotify, YouTube, etc.). Returns matching episodes with their id, title, podcast, podcast_id, and date. Use the id with get_transcript, or the podcast_id with list_episodes to get the show's whole back-catalog. Does not consume credits.
query必填Get transcript
Fetch a podcast episode's transcript as clean Markdown with real speaker names and timestamps. Pass an episode id from search_podcasts. Costs 1 credit on the first fetch of an episode; repeat fetches are free and errors are never charged.
episode_id必填List a show's episodes
List a podcast's entire back-catalog (every episode, newest first). Pass a podcast_id from a search_podcasts result. Returns each episode's id, title, and date — fetch any with get_transcript. Use this to transcribe a whole show. Does not consume credits itself; transcribing the returned episodes costs 1 credit each (repeat fetches are free), so make sure the key has enough credits before looping.
podcast_id必填Get credit balance
Check the current Spoken credit balance, account email, and recent usage for the configured API key. Does not consume credits.
無需輸入參數。
List followed shows
The shows this key is kept current on. Every charged fetch makes its show a follow: the five most-fetched shows from the last 180 days are followed automatically (source 'fetch'), and up to 25 more can be declared with follow_podcast (source 'explicit'). Muted shows are listed separately. Use list_new_episodes to see what is new on them. Does not consume credits; the demo key has nothing to follow with.
無需輸入參數。
Follow a show
Declare a follow for a show, or clear a mute. Pass a podcast_id from search_podcasts. A declared follow stays until unfollow_podcast, regardless of fetch activity, and does not use one of the five automatic slots. For a show never fetched, what counts as new starts from its newest episode now. Does not consume credits.
podcast_id必填Stop following a show
Mute a show: it leaves the followed list and later fetches do not re-add it. follow_podcast reverses it. Pass a podcast_id from list_following. Does not consume credits.
podcast_id必填List new episodes on followed shows
What is new on the shows this key follows: for each show, the episodes released in the last 90 days that are newer than the newest one already fetched from it (up to 10 per show), each with a transcript_url. Episodes with no transcript are left out. Fetch any with get_transcript (1 credit each on first fetch); a fetch raises that show's floor, so the next call lists only what came after. Refreshed hourly, so a show followed moments ago may be empty until its first poll. Does not consume credits.
無需輸入參數。
已滿足適用檢查項的 100%
該評分衡量此收錄條目的完整程度，依據註冊中心條目、端點與公開來源計算得出。它不是對伺服器本身的評價，也不表示伺服器安全或目前可用：可連通性是帶日期的檢查結果，評分也不包含安裝量或使用者評價。
# 1. Find an episode (by text, or paste a Spotify/YouTube URL)curl -s 'https://spoken.md/search?q=huberman+sleep' \ -H 'x-api-key: pt_demo'
# 2. Fetch the transcript as Markdowncurl -s 'https://spoken.md/transcripts/1000651996090' \ -H 'x-api-key: pt_demo'**John Smith** (0:00)Welcome to the show. Today we're talking about...
**Jane Doe** (0:15)Thanks for having me.pip install spoken-mdfrom spoken_md import Spoken
spoken = Spoken() # SPOKEN_API_KEY, or the demo keyepisode = spoken.search("huberman sleep")[0]print(spoken.transcript(episode.id)) # Markdown with real speaker namesclaude mcp add --transport http spoken https://spoken.md/mcp{ "mcpServers": { "spoken": { "command": "npx", "args": ["-y", "spoken-mcp"], "env": { "SPOKEN_API_KEY": "pt_your_key" } } }}npm install && npm run buildSPOKEN_API_KEY=pt_your_key node dist/index.js