[Proposal] PodcastIndex MCP Server - Authoritative AI-Powered Podcast Discovery for the Open Ecosystem

[cclawton]

Summary
I've built (well Claude built) an open-source Model Context Protocol (MCP) server for PodcastIndex that enables AI assistants to search podcasts, retrieve episodes, and query Value4Value data. I believe this aligns strongly with the Podcasting 2.0 mission and wanted to share it with the community for feedback and potential adoption.
Repository: https://github.com/cclawton/podcastindex-mcp-server
Why This Matters for Podcasting 2.0

Just as PodcastIndex provides an open alternative to closed podcast directories, this MCP server ensures AI-powered podcast discovery remains: Decentralized, Censorship-resistant, Community-owned

The AI landscape is rapidly consolidating around a few major players. Without an open MCP server pointing to PodcastIndex, AI assistants will default to whatever closed, ad-driven, or editorially-curated sources their vendors choose. PodcastIndex should be THE authoritative source for AI podcast discovery. Imagine Podcasting 2.0 apps like Fountain, CurioCaster, or Podverse integrating MCP to let users:

Ask "Find podcasts about Bitcoin Lightning" and get results from the open index
Query "What new shows launched this week in the Technology category?"
Discover Value4Value-enabled shows through natural language

This extends the namespace vision into the AI era.
What It Does
The MCP server exposes PodcastIndex API functionality to any MCP-compatible AI client:
Search: podcast_search_byterm, podcast_search_bytitle, podcast_search_byperson, podcast_search_music
Podcast Data: podcast_get_byfeedid, podcast_get_byfeedurl, podcast_get_byguid, podcast_get_byitunesid
Episodes: episode_get_byfeedid, episode_get_byid, episode_get_byguid, episode_get_live, episode_get_random
Discovery: podcast_get_trending, recent_get_episodes, recent_get_feeds, recent_get_newfeeds
Value4Value: value_get_byfeedid, value_get_byfeedurl, value_get_bypodcastguid
Metadata: stats_get_current, categories_list, hub_pubnotify
Real-World Example
Query: "Find podcast episodes featuring David Deutsch"
Result via MCP:
Found 18 episodes featuring David Deutsch including:

"David Deutsch Says We Will Build Humans Before We Build AGI" - Into the Impossible (33 min)
"AGI, quantum computing, and the future of humanity" - Strange Loop Podcast (58 min)
"Science, Complexity, and Explanation" - Sean Carroll's Mindscape (102 min)
"Disagreeing about Belief, Probability, and Truth" - Increments Podcast (92 min)
"The Universal Constructor" - Fallible Animals (26 min)

This query executed in seconds, returning results from the open PodcastIndex - not a curated, filtered, or censored corporate directory.
AI Platform Compatibility
The MCP server works with:

Claude Desktop - Tested and working
Any MCP Client (including of course Podcast 2.0 apps, granted it would need an LLM or local SLM to work) - Standards-compliant

Built With
This server was built using claude-flow (https://github.com/anthropics/claude-flow) - an agentic development framework - demonstrating how AI can help build tools for AI. The entire development process, from API mapping to error handling, was AI-assisted while maintaining human oversight and code ownership.
Questions for the Community

Official adoption? Would PodcastIndex consider maintaining an official MCP server, similar to how you maintain the API docs?
Namespace extension? Could MCP endpoints become part of the Podcasting 2.0 namespace vision - for example, a podcast:mcp tag pointing apps to discovery services?
V4V for AI? Could AI queries eventually support Value4Value - streaming sats to index maintainers when AI assistants use the service?
App integration? Are any Podcasting 2.0 app developers interested in integrating MCP for AI-powered discovery?

Next Steps
I'm happy to:

Transfer ownership to Podcastindex-org if there's interest

The code is MIT licensed and ready for community use.
Let's keep podcast discovery open - even in the age of AI.
Craig Lawton (craig.lawton@protonmail.com)
GitHub: https://github.com/cclawton
LinkedIn: https://www.linkedin.com/in/craiglawton

[adamc199]

Can't wait to try this out!

[daveajones]

I’m on board with this and thank you for doing it! Ill fork this repo over to podcastindex-org and see if I can get it working to start testing.

[cclawton]

Had a crack at swarming a sample podcast app to show intended functionality of MCP. Sample code here https://github.com/cclawton/podcast-2-0-ai-player/blob/main/README.md https://www.linkedin.com/posts/craiglawton_ai-softwaredevelopment-opensource-activity-7420352978391580672-ZOmT?utm_source=share&utm_medium=member_desktop&rcm=ACoAAACIBe4BJInQQa5_rtJ3J41EziV7gCaMzW8 Sent with [Proton Mail](https://proton.me/mail/home) secure email.

…

On Thursday, 15 January 2026 at 23:20, Dave Jones ***@***.***> wrote: daveajones left a comment [(Podcastindex-org/docs-api#151)](#151 (comment)) I’m on board with this and thank you for doing it! Ill fork this repo over to podcastindex-org and see if I can get it working to start testing. — Reply to this email directly, [view it on GitHub](#151 (comment)), or [unsubscribe](https://github.com/notifications/unsubscribe-auth/AEHJT72LIKXZZPQB3LPFJYT4G6A2FAVCNFSM6AAAAACRXOE26SVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZTONJUGQ3DEMJQHA). You are receiving this because you authored the thread.Message ID: ***@***.***>

[daveajones]

This is great! I've been struggling to get the time to put testing together on this. I'll mention it in the board meeting today and see if we can get more testers on board.

Forgive my ignorance - is there a potential for the MCP server to swamp the Index API with calls if the prompts are broad? For instance, in your "Find podcast episodes featuring David Deutsch" example, would that potentially translate to the MCP server trying to crawl the whole index (like "grepping" the api) under certain contexts?

[cclawton]

It's a good question, and you made me go double check. The MCP implementation translates by relying on server-side search via Podcast Index API endpoints, not client-side crawling. It translates natural language into one of byperson (for David Deutsch style queries), byterm and bytitle API queries. One MCP query to one API query. Someone could write client-side logic that pulls a long list of podcasts, then queries every episode for all these podcasts and greps results, but they could do that without MCP now, and it looks like the index has server side throttling to protect against this. So I can't on the surface of it see an issue, but monitoring and server-side throttles would protect. As a potential future enhancement, you could create a podcastindex knowledgebase/search endpoint for more general semantic style queries, potentially enhanced with transcripts, but I'm getting ahead of myself. Sent with [Proton Mail](https://proton.me/mail/home) secure email.

…

On Saturday, 24 January 2026 at 04:14, Dave Jones ***@***.***> wrote: daveajones left a comment [(Podcastindex-org/docs-api#151)](#151 (comment)) This is great! I've been struggling to get the time to put testing together on this. I'll mention it in the board meeting today and see if we can get more testers on board. Forgive my ignorance - is there a potential for the MCP server to swamp the Index API with calls if the prompts are broad? For instance, in your "Find podcast episodes featuring David Deutsch" example, would that potentially translate to the MCP server trying to crawl the whole index (like "grepping" the api) under certain contexts? — Reply to this email directly, [view it on GitHub](#151 (comment)), or [unsubscribe](https://github.com/notifications/unsubscribe-auth/AEHJT74FREGKGROBAMTMSCL4IJJI5AVCNFSM6AAAAACRXOE26SVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZTOOJRGM2DEMJXHE). You are receiving this because you authored the thread.Message ID: ***@***.***>