You mentioned the idea of having an agent read the show RSS feed for summaries and stuff, but with the size of the RSS data, that would be a lot of tokens to process, so I'd recommend using MCP tools to process and search the data first. Here's an MCP server that I wrote to handle retrieving episode and transcript data from Podcasting 2.0 compatible RSS feeds. I've really only tested it with JB feeds so far though: https://github.com/Red5d/podcast_mcp