Spoken.md

Spoken.md turns a published podcast episode into clean Markdown in one API call — with real speaker names, not "Speaker 1". It's a retrieval API, not a speech-to-text service. Spoken works on shows that already publish a transcript and returns that text as structured Markdown with named, timestamped speakers — ready to drop straight into an LLM context window, a RAG pipeline, or a research corpus. Nothing is uploaded, nothing sits in a queue, and there's no per-minute transcription bill: a three-hour episode costs exactly the same as a ten-minute one. If an episode has no transcript available, the API returns 404 and you aren't charged. Built for agents as much as for people. There's a documented REST API, an OpenAPI spec, llms.txt and agents.md, a published Agent Skill, and an MCP server (npx -y spoken-mcp) so an assistant can search and fetch transcripts on its own. Search by keyword or paste an episode URL, or enumerate a show's back catalogue to archive it in bulk. Pricing is pre-paid credits, no subscription: $15 for 100 transcripts ($0.15 each), $50 for 500 ($0.10), or $160 for 2,000 ($0.08). One credit per successful fetch, re-fetching an episode you already pulled is free, and credits never expire. A free demo key returns a sample episode so you can see the exact output before paying anything. Best for developers, researchers and analysts building a searchable archive of the shows they actually follow.