Podcast Robot

Purpose

A zero-setup client for the Podcast Robot MCP server: search podcasts, check what's new in followed shows, find who was on a show, shows near a place, search episode transcripts, make shareable clip references, list live streams, read Podping events, grade a podcaster's feed. No API key, no account, no OAuth. The server is public and stores nothing about the user. Full guide: https://podcastindexrobot.com/llms.txt

Tooling

bin/mcp_robot.py is a minimal JSON-RPC client (Streamable HTTP via curl).

List available tools:

bin/mcp_robot.py tools

Call a tool (args as JSON):

bin/mcp_robot.py call <tool> '<json-args>'

Start here (a live menu plus an onboarding checklist):

bin/mcp_robot.py call start_here '{}'

Search for a show:

bin/mcp_robot.py call search_podcasts '{"query": "No Agenda", "mode": "title", "max_results": 3}'

Find a moment in a transcript (needs the episode id from get_episodes / whats_new), then clip it:

bin/mcp_robot.py call get_episodes '{"feed": "41504", "max_results": 1}'
bin/mcp_robot.py call transcript_search '{"episode": "61124290178", "query": "bitcoin", "max_hits": 2}'
bin/mcp_robot.py call make_clip '{"episode": "61124290178", "start": 4577, "duration": 45}'

Preferences (the server is stateless)

Ask the user once for their podcast app, city, the shows they follow and topics they like; keep the answers in your own memory and pass them as prefs on every call, e.g. {"prefs": {"app": "fountain", "city": "Austin", "follows": ["41504"], "language": "en"}}. With prefs.app set, every show and episode result carries open.app, a link that opens in the user's app. Without it, use open.universal.

Auth

None. The endpoint https://podcastindexrobot.com/mcp is open. Do not ask the user for credentials and do not invent any.

Operating Rules

1. Be polite to a shared hosted endpoint: the limit is 60 calls per minute per IP. When keeping watch, pass the time of your last check as since (whats_new, podping_events, feed_activity) and check at most every 15 minutes.
2. transcript_search returns media-fragment play URLs (#t=start,end); use make_clip for a shareable clip reference.
3. podping_events with resolve: true adds feedId and title to each event; contains filters feed URLs.
4. Every result ends with hints: suggested next calls. Offer the user two or three of them.
5. Results with "kind": "fault" carry an errorId: tell the user it is logged and being fixed, quote the id, retry once. Results with "kind": "input" are not bugs: follow whatToTry, closestMatches or howToAdd.
6. Never present invented episode details; everything comes from tool output.