Show HN: 通过 Claude 的 MCP 搜索播客、法庭听证会和财报电话会议

1作者: lukaesch9 个月前
我一直在研究一种方法,将 Claude(以及其他模型上下文协议客户端,如 ChatGPT)连接到大型的公共和用户上传音频索引。 结果是 Audioscrape 的一个远程 MCP 服务器,它允许 AI 助手查询并返回以下内容的关联片段: * 热门播客 * 法庭听证会 * 财报电话会议 * 您自己上传的任何音频 示例查询: “Joe Rogan 在他的播客中对德国说了什么?” MCP 服务器处理转录索引并公开一个 search\_audio\_content 工具。然后,Claude 调用它并从音频中提取精确匹配的时刻。 技术说明: * 服务器:远程 MCP 规范实现 * 搜索:转录文本的全文索引 * 客户端:与 Claude 桌面版、移动版、Web 版 MCP 支持兼容 * 数据:公共音频源 + 用户上传的组合 文档和设置: * Claude 设置:[https://www.audioscrape.com/docs/claude-setup](https://www.audioscrape.com/docs/claude-setup) * MCP 服务器详细信息:[https://www.audioscrape.com/docs/mcp](https://www.audioscrape.com/docs/mcp) 我很好奇其他人如何扩展 MCP 来搜索非文本数据源——特别是如果您尝试将 AI 助手与大型数据集(超出常用 API)集成。
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>&quot;What is Joe Rogan saying about Germany on his podcasts?&quot;<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 &amp; setup:<p>Claude setup: <a href="https:&#x2F;&#x2F;www.audioscrape.com&#x2F;docs&#x2F;claude-setup" rel="nofollow">https:&#x2F;&#x2F;www.audioscrape.com&#x2F;docs&#x2F;claude-setup</a><p>MCP server details: <a href="https:&#x2F;&#x2F;www.audioscrape.com&#x2F;docs&#x2F;mcp" rel="nofollow">https:&#x2F;&#x2F;www.audioscrape.com&#x2F;docs&#x2F;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.