Spoken is an API for people who follow a few podcasts closely and want the episodes as text they can search, quote and hand to an LLM.
Two calls do the job. One lists every episode of a show that has a transcript. One returns an episode as Markdown, with the speakers named and each turn timestamped. Every show page on spoken.md carries a bash loop that pulls a whole back catalog; 215 episodes of Acquired come to about $32.
What it is not
Spoken is not speech-to-text. It serves transcripts of published podcast episodes, so it is priced per episode rather than per minute of audio, and a show that publishes no transcript is one it cannot serve. Search yours first; every result it shows is fetchable.
Pricing
Pre-paid credits, no subscription. $15 for 100 episodes, $50 for 500, $160 for 2,000. Credits never expire, an episode that cannot be served is never charged, and re-fetching one you already bought is free. A demo key returns one full episode with no signup.
For agents
OpenAPI spec, MCP server (npx spoken-mcp), Python client (pip install spoken-md), a Claude Code and Cursor skill, and llms.txt.
