展示HN：通过Claude的MCP搜索播客、法庭听证会和财报电话会议
我一直在研究如何将Claude（以及其他模型上下文协议客户端，如ChatGPT）连接到一个大型的公共和用户上传音频索引。<p>结果是为Audioscrape开发的远程MCP服务器，它允许AI助手查询并返回以下内容的相关片段：<p>热门播客<p>法庭听证会<p>财报电话会议<p>任何您自己上传的音频<p>示例查询：<p>“乔·罗根在他的播客中谈论德国的内容是什么？”<p>MCP服务器处理转录索引，并提供一个search_audio_content工具。然后，Claude调用它并提取音频中的确切匹配时刻。<p>技术说明：<p>服务器：远程MCP规范实现<p>搜索：对转录文本的全文索引<p>客户端：支持Claude桌面版、移动版和网页MCP<p>数据：公共音频源与用户上传的组合<p>文档与设置：<p>Claude设置：<a href="https://www.audioscrape.com/docs/claude-setup" rel="nofollow">https://www.audioscrape.com/docs/claude-setup</a><p>MCP服务器详情：<a href="https://www.audioscrape.com/docs/mcp" rel="nofollow">https://www.audioscrape.com/docs/mcp</a><p>我很好奇其他人如何扩展MCP以搜索非文本数据源，特别是如果您尝试过将AI助手与常规API之外的大型数据集集成的话。
查看原文
I’ve been working on a way to connect Claude (and other Model Context Protocol clients like ChatGPT) to a large index of public and user-uploaded audio.<p>The result is a Remote MCP server for Audioscrape that lets an AI assistant query and return relevant segments from:<p>Popular podcasts<p>Court hearings<p>Earnings calls<p>Any audio you upload yourself<p>Example query:<p>"What is Joe Rogan saying about Germany on his podcasts?"<p>The MCP server handles the transcription index and exposes a search_audio_content tool. Claude then calls it and surfaces exact matching moments from the audio.<p>Tech notes:<p>Server: Remote MCP spec implementation<p>Search: Full-text index over transcripts<p>Client: Works with Claude Desktop, Mobile, Web MCP support<p>Data: Combination of public audio sources + user uploads<p>Docs & setup:<p>Claude setup: <a href="https://www.audioscrape.com/docs/claude-setup" rel="nofollow">https://www.audioscrape.com/docs/claude-setup</a><p>MCP server details: <a href="https://www.audioscrape.com/docs/mcp" rel="nofollow">https://www.audioscrape.com/docs/mcp</a><p>I’m curious how others might extend MCP for searching non-text data sources - especially if you’ve tried integrating AI assistants with large datasets outside the usual APIs.