Make your AI assistant smart about podcasts. Connect it in one minute.
Podcast Robot is a plug-in for AI assistants like Claude, ChatGPT or your own agent. Once connected, your assistant can see every podcast in the world through Podcast Index, the open directory of over 4 million shows, plus the live signals that say when something new drops or a show goes live. No account, no key, nothing to install on your phone. Your podcast app stays your podcast app: Podcast Robot hands you links that open right in it.
Just talk to your assistant the way you would to a friend who listens to everything:
Many assistants can run a request on a schedule. Try one of these, in your own words:
Podcast Robot tells your assistant how to check cheaply (only what changed since last time), so one check costs one or two calls. Honest note: hosted assistants schedule hourly at best (ChatGPT, Claude Routines, Dots) or daily (Gemini Spark, Grok, Manus); a true "every 15 minutes" needs an agent you run yourself, such as Muse or OpenClaw. Each entry in the connect list says what it can do.
Podcasters: ask "grade my feed" and Podcast Robot tells you which Podcasting 2.0 tags you are missing and what each one would let assistants do for your listeners.
Copy this address. Then follow the steps for your assistant.
https://podcastindexrobot.com/mcp
"Plan" says what the app requires. Checked October 2026; apps change this often.
Meta Muse app (phone or web): not yet. It only uses connectors that Meta has reviewed, and it will refuse a request to use an unknown server, which is the right thing for it to do. We are asking Meta about a reviewed Podcast Robot connector. Until then, use Claude, Le Chat, Gemini or Grok on your phone, or one of the agents below.
Muse Code, OpenClaw and other agents that can run code: send this message, or add the server in their settings:
The guide your agent reads is https://podcastindexrobot.com/llms.txt. Podcast Robot is a project of Podcast Index, the open podcast directory; it stores nothing about you and needs no login.
Phone app: wherever you chat with such an agent, including your phone.
Schedules: an agent that runs jobs on a timer can check every 15 minutes; it only asks what changed since last time.
Plan: Free (one custom connector allowed), Pro, Max, Team, Enterprise.
Phone app: yes. Anthropic: connectors are available on Claude for iOS and Android; add it on the web or desktop first (adding from the phone is in beta), then it works in the phone app.
Schedules: Claude Routines (paid plans) run hourly, daily or weekly with your connectors; Pro 5 runs a day, Max 15.
Plan: Free and paid. On team accounts an admin adds it.
Phone app: yes - the Le Chat phone app has the same connectors as the web.
Schedules: yes, scheduled tasks in Work mode use your connectors (one-time or recurring).
Plan: a personal Google account (not work or school), United States, 18 or over, Keep Activity ON (custom apps are unavailable when it is off), English only.
Phone app: yes - Google: custom apps can only be connected in the Gemini web app, and "after it's connected, you can use the custom connected app in both the Gemini mobile app and Gemini web app."
Schedules: through Spark (Google AI Pro or Ultra): recurring schedules on a cadence you set; custom apps work only inside Spark tasks.
Plan: all bring-your-own MCP connectors need SuperGrok (standalone or via X Premium+).
Phone app: probably yes - reports say connectors reached the iOS and Android apps in May 2026; add it on grok.com first.
Schedules: Grok Tasks / Automations run daily, on weekdays, weekly, monthly or a custom cadence.
Plan: Pro, Max or Enterprise. Not on Free.
Phone app: yes - custom connectors are account-scoped: add it on the web once and it works in the iOS, Android, macOS, Windows and web apps on the same account.
Schedules: yes, Scheduled Tasks and Automations in Perplexity Computer can use your connectors.
Plan (OpenAI Help Center, checked October 2026): Pro can connect custom MCP servers with read/fetch tools in Developer mode, which is all Podcast Robot needs; Business, Enterprise and Edu get full MCP (an admin enables Developer mode). Plus: OpenAI's help article does not list Plus, but its developer guide does. It may work on the web; try it. Free: no.
Phone app: no. OpenAI: "Are MCP apps available on mobile? No - web only." Setup and use are on chatgpt.com in a browser. On a phone, use Claude, Grok, Gemini or Le Chat instead; they all work in their phone apps.
Schedules: scheduled tasks can use connected apps on the web, at most once an hour.
A dot "can use supported plugins installed and enabled for your account" (Gmail, Drive, GitHub and the like). OpenAI's Dots documentation does not mention custom MCP connectors. Best try: add Podcast Robot to ChatGPT as above, then see whether your dot lists it under its apps. We will update this line when OpenAI does.
Phone app: no - a dot's apps come from ChatGPT's connectors, which are web-only.
Schedules: dots are always-on and run recurring tasks, up to hourly on paid plans.
Plan: Dots themselves need ChatGPT Pro 100 / 200 / 500 (18+, not in the EEA, UK or Switzerland), Business Premium or Enterprise; rolling out gradually.
Grok Bot is not the Grok chat above: it is a team of always-on agents, each on its own cloud computer with a browser and a terminal. Two ways in:
Plan: SuperGrok Heavy, Cursor Ultra or Cursor Teams Premium (beta since August 2026; desktop and iOS).
Phone app: yes on iOS (Android planned); the connector is added at grok.com.
Schedules: always-on; routines run work on a schedule or on events, so it can keep watch.
Plan: Manus is a paid product; its docs do not say which tier includes custom servers.
Phone app: not confirmed - add it on the web; the docs do not say whether custom servers work in the phone app.
Schedules: Scheduled Tasks 2.0, daily, weekly or monthly.
Fastest: install the published skill from ClawHub, then just talk about podcasts:
clawhub install podcast-robot · clawhub.ai/skills/podcast-robot
Or add the server by hand:
Plan: free, you run it yourself.
Phone app: yes - you talk to your agent from any chat app it is wired to (Telegram, Signal, WhatsApp...).
Schedules: built-in cron, any interval; every 15 minutes is fine.
The consumer Copilot app has no place to add a custom MCP server. Microsoft 365 Copilot can reach MCP servers, but only when a company admin adds them (federated connectors, Copilot Studio).
Download the Podcast Robot skill (a SKILL.md plus a tiny Python client, written by Adam's Muse and tidied by us) and drop the podcast-robot folder into your agent's skills directory. OpenClaw users: clawhub install podcast-robot (ClawHub page). The agent then knows when and how to use Podcast Robot, needs no key, and keeps your preferences itself.
claude mcp add --transport http podcast-robot https://podcastindexrobot.com/mcphttps://podcastindexrobot.com/mcp, no auth.Free during the test. Please don't let scripts hammer it: 60 requests a minute per connection is the limit. Terms · Privacy
We record which tools are used and the search words, with no IP addresses and nothing that identifies you, only to improve Podcast Robot. Records are deleted after 30 days.
Remote MCP server, Streamable HTTP, stateless, JSON responses, no auth. Endpoint https://podcastindexrobot.com/mcp.
Data: the Podcast Index API, each show's RSS feed (Podcasting 2.0 namespace), Podping on Hive (real time), and the Podcast Index public datasets (locations, copy-cat feeds,
podroll recommendations, hosting stats, index counts). Nothing is stored about users: no accounts, no history.
Every tool accepts prefs: {"app": "fountain", "city": "Austin", "follows": ["41504"], "topics": ["bitcoin"], "language": "en", "clean_only": false, "v4v_only": false}.
The agent keeps these in its own memory. With prefs.app set, every show and episode carries open.app (a link in the owner's app),
open.universal (a page like /open/41504 offering every app) and open.podcastindex.
| App | Show link key | Episode deep link |
|---|---|---|
| Fountain | Podcast Index id | yes |
| Castamatic | podcast GUID | no, opens the show |
| Podverse | Podcast Index id | no, opens the show |
| CurioCaster | Podcast Index id | no, opens the show |
| Podcast Guru | Apple id | no, opens the show |
| Apple Podcasts | Apple id | no, opens the show |
| Overcast | Apple id | no, opens the show |
| Pocket Casts | Apple id | no, opens the show |
| Castro | Apple id | no, opens the show |
| AntennaPod | feed URL | no, opens the show |
| Podcast Addict | feed URL | yes |
| Breez | feed URL | yes |
| TrueFans | podcast GUID | no, opens the show |
| Podfriend | Apple id | yes |
| Steno.fm | podcast GUID | yes |
| Dial (Houston's app) | yes | |
| Spotify | none (closed ids) | none |
| YouTube Music | feed URL | no, opens the show |
Free during the test. Limit: 60 requests per minute per IP (burst 30); over that you get HTTP 429 with Retry-After. Please do not poll in tight loops: Podping events arrive within seconds, so once a minute is plenty.
We record which tools are used and the search words, with no IP addresses and nothing that identifies you, only to improve Podcast Robot. Records are deleted after 30 days.
start_here | START HERE. Call this the first time podcasts come up, or when the user asks what you can do. Returns a short live menu built from real data right now: a show that is live at this moment, what is trending, a fresh clip published today, a show made near the user (pass `place`, e.g. their city, if you know it), the newest Value4Value show, and the podcast that was updated seconds ago on Podping. Offer the user 2-3 of these as concrete things to try, then keep the rest for later. |
|---|---|
get_clips | Ready-made clips and structure for an episode: the podcaster's own `soundbites` (as playable clip references with `playUrl`, start and duration) and the `chapters` list with timestamps. Also says whether a transcript exists so `transcript_search` can cut custom clips. `episode` = Podcast Index episodeId or episode guid (+ `podcast_guid`). |
transcript_search | Find where an episode talks about something. Reads the episode's Podcasting 2.0 transcript and returns the best-matching passages with start/end times, the words said, and a `playUrl` that jumps straight to that moment. Use for "find the part where...", quoting a host with a timestamp, or picking the start of a clip for `make_clip`. |
make_clip | Build a shareable clip reference: `shareUrl`, a public page on podcastindexrobot.com with the art, the transcript words of the window, a player that plays only that window and open-in-app links (send THIS to people; it previews in iMessage/WhatsApp/X); plus the raw `playUrl` (`#t=start,end` media fragment), the audio URL, start, duration and the transcript text. Clips are capped at 5 minutes. No audio is cut server-side. |
episode_outline | What an episode is about, cheaply: the podcaster's own chapters (podcast:chapters) as an outline with timestamps, titles and links, plus a one-line `about` built from the chapter titles. A chapters file is a few KB, against 100 KB-2 MB for a transcript, so call this FIRST for "what did they talk about", "summarize", "which part is about X". Each chapter gives the timestamp to then quote or clip just that window with `transcript_search` / `make_clip`. Episodes without chapters say so and point to `transcript_search`. `episode` = Podcast Index episodeId or episode guid (+ `podcast_guid`). |
chapter_search | Which episodes of a show talked about something, answered from chapter titles only (no transcripts): reads the podcast:chapters files of the show's last `max_episodes` episodes (max 50) and returns the episodes whose chapter titles contain the query words, with the matching chapters and their timestamps. Cheap and fast; use it before any transcript search when the question is "when did they talk about X". `feed` = feedId, podcastGuid, feed URL or show name. |
feed_tags | The full Podcasting 2.0 view of a show, read live from its feed: hosts and regular cast (`persons`), `location`, `podroll` (shows the host recommends), `funding`, `value` (how to support it), `liveItems` (scheduled / live streams with stream URL and chat), `updateFrequency`, `license`, `trailers`, `social`, and per-episode tag coverage. The first `max_items` episodes come back with their persons, soundbites, transcripts, chapters and socialInteract links. |
podroll_walk | Follow a show's `podroll` — the shows its creator recommends — and the podrolls of those shows, up to `hops` deep (max 2). This is discovery by trust: a human picked each link. Returns each hop with compact show records and who recommended them. Shows without a podroll end the walk on that branch. |
podcasts_near | Find shows by place, using every feed in the index that carries a `podcast:location` tag (about 18,000 feeds, refreshed daily from the Podcast Index public dataset). |
recommendations | The podroll graph around a show, from the index-wide recommendation dataset (every `podcast:podroll` in the index, mined daily): `recommendedBy` = shows whose creators put this show in their podroll (reverse podroll — who vouches for it, with a popularity count), and `recommends` = the shows this one lists. No feed fetch needed, so it is fast; `podroll_walk` reads the live feed and walks deeper. |
most_recommended | The shows most often placed in other podcasters' podrolls — a human-curated popularity chart, not download numbers. Optional `medium` filter ('podcast', 'music', 'video'). |
grade_feed | Ask Engine Optimization report for a podcaster: grades the feed's Podcasting 2.0 tag use (0-100, A-F), lists what is present, partial and missing, and explains — in terms of what an AI agent could then do for listeners — why each missing tag matters and what to add first. |
index_stats | How big is podcasting, right now: Podcast Index totals (feeds, episodes), how many feeds published in the last 3 / 7 / 30 / 90 days, how many new episodes, Podcasting 2.0 tag adoption (value blocks, funding, transcripts, chapters, soundbites, socialInteract, music and video mediums), the top hosting companies by live feed count, and Podping/dataset health of this server. Good for "how many podcasts are there?", "how many are actually active?", and for putting a feed grade in context. |
add_to_index | Add a podcast feed to Podcast Index (or ask the index to re-read it now). ONLY call this when the owner explicitly asked to add their feed - it notifies the index on their behalf. |
copycat_feeds | Copy-cat and lookalike feeds for a show, from the Podcast Index lookalike detector (daily export): every pair of feeds that share this show's episodes, which side the detector judges to be the original, a label, a confidence 0-1 and the signals in plain words. Use it when a listener asks "is this the real feed?", when a podcaster asks "is someone copying my show?", or to pick the canonical feed among several with the same name. |
copycat_hosts | Who copies whom, by hosting company: counts of lookalike pairs between an original's host and the lookalike's host, from the Podcast Index lookalike detector (daily export). Default `label='copy'` = feeds re-hosting another show's episodes. Give `host` (e.g. 'spreaker.com', 'anchor.fm') to see one host on either side; leave it empty for the index-wide top list. Use it for "which platforms carry the most copied shows?" or for a host that wants to know what the detector sees on its side. Counts are the detector's likely classifications, not determinations. |
search_podcasts | Find podcasts (or music / video / audiobook / film feeds) in Podcast Index. |
search_people | Find episodes where a person appears, by name — across every show in the index. |
get_podcast | Get one show by Podcast Index `feedId`, `podcastGuid`, feed URL, or `itunes:<id>`. |
get_episodes | List episodes of a show, newest first. `feed` = feedId, podcastGuid or feed URL. |
get_episode | Get one episode with full show notes by Podcast Index `episodeId` or episode `guid`. |
trending | What is hot right now across the index, by trend score. |
surprise_me | Serendipity: random episodes from the whole index, optionally filtered by language and category. Each call returns different results. Use it for "surprise me", "something I would never find on my own", or to seed a listening session. Pair with `get_podcast` to tell the user about the show, and `value_info` if they like it and want to support it. |
recent | The newest things in the index, in reverse chronological order. |
live_now | Shows that are streaming LIVE right now (Podcasting 2.0 `liveItem`), with the stream link. Ask again in a few minutes; live status changes fast. |
whats_new | "What's new for me?" — for a list of shows the user follows, return new episodes since a time, newest first, plus the show's current live status. `feeds` are feedIds, podcastGuids or feed URLs (mix allowed). `since`: ISO 8601, epoch seconds, or '7d' / '36h'. |
value_info | How to support a show: its Podcasting 2.0 value block (Lightning / Value4Value recipients and splits, suggested amount) and funding links. Pass `episode_guid` for an episode-level block (some shows split differently per episode). `feed` = feedId, podcastGuid or URL. |
podping_events | Real-time feed activity from Podping: which feeds announced an `update` (new episode or change), went `live`, or ended a live stream (`liveEnd`), since a time. Updates arrive within seconds of publishing — before any crawler. `since` like '15m', '2h', ISO 8601 or epoch; the server keeps up to 48 h (see `coverageFrom`). `medium` filters e.g. 'podcast', 'music', 'podcastL' (live podcasts) or 'live' for all live mediums. `resolve=True` looks up each feed in Podcast Index (slower; use with small `max_results`). |
feed_activity | For specific shows: the Podping events (update / live / liveEnd) since a time, resolved through Podcast Index so you can pass feedIds, podcastGuids or URLs. Tells you *that* a feed changed before `get_episodes` can show *what* changed. Combine with `whats_new`. |
Start here: what can Podcast Robot do? | A live menu of things to try right now |
|---|---|
Surprise me | Random, serendipitous episodes with a reason to listen |
What's new for me | New episodes and live streams for the shows you follow |
Podcasts near me | Shows made in, or about, a place |
Grade my feed | Ask Engine Optimization: which Podcasting 2.0 tags to add and why |
Clip this moment | Find where an episode talks about something and get a jump link |
Who is this person? | Everywhere a host or guest appears across the index |
Walk the podroll | Discovery by trust: the shows a host recommends, two hops out |
Discover podcasts about a topic | Search, trending and a surprise pick |
pi://categories | Podcast Index category names and ids |
|---|---|
pi://stats | Podcast Index live statistics |
pi://podping/status | Podping watcher health and coverage window |
pi://datasets/status | Public dataset (locations, recommendations) load status |
Support · Privacy policy · Terms of service · README · health · source patterns for app links: podcast-platform-links · version 0.1.0