About ourword
ourword is an MCP server for searching Chinese deep reads of more than 190 Chinese- and English-language podcasts. It provides searchable episode entries and timestamped points, short quotes, and figures, but does not provide transcripts.
Key features
- 01Search episodes by keyword, show, or category
- 02Retrieve timestamped points, quotes, and figures
- 03Browse shows and their deep-read update details
- 04List recent deep reads by category
Use cases
- 01Find podcast episodes discussing a topic
- 02Check the audio timestamp for a point or quote
- 03Follow the latest deep reads from shows
MCP support
The MCP server provides tools to search episodes, retrieve structured episode deep reads, list shows, and browse recent entries; it does not provide transcripts.
Configuration
Local server config is the same for every client. Copy it into your client’s MCP config file.
We haven’t collected a package or remote endpoint for this server. Follow the instructions in the project README.
Tools
Local servers expose tools only once running on your machine, and the maintainer has not provided a tool list.
Details
From the project READMEourword MCP server
把 原声(ourword.ai)的播客深读开放给 agent 查。
每天从 190 多档中英文播客里挑出值得记住的判断,写成中文深读。要点、金句、数字 都带时间戳,点一下回到它在原声里被说出的那一秒 —— 这个服务器把同一批数据变成 可查的接口。
为什么需要它
站上本来就有 llms.txt / llms-full.txt,但那是给「整篇读一遍」准备的:
llms-full.txt 有 6.9 MB,agent 要回答「谁讲过 reward hacking」只能把它整份读进
上下文。这个服务器让它先筛出几条,再按需要取那一集的结构。
装
不用装。只要有 Python 3.10+:
{
"mcpServers": {
"ourword": {
"command": "python3",
"args": ["/path/to/mcp/ourword_mcp.py"]
}
}
}
只用标准库,没有第三方依赖。数据从线上取,所以不会拿到发行当天的旧快照。
四个工具
| 工具 | 做什么 |
|---|---|
search_episodes | 按关键词、节目、分类找集 |
get_episode | 一集的要点/金句/数字/术语,每条带时间戳 |
list_shows | 在册节目,各自已深读多少篇、最新一篇是哪天 |
latest_episodes | 最近的深读,可按分类 |
关于原文
这里不提供逐字稿。 逐字稿是各播客的版权内容,站上任何一份文件里都没有, 这个接口也不会有。给出去的是我们自己写的判断、可核对的数字、带署名的短引用, 以及回到原音频那一秒的链接 —— 引用的人可以自己去核。
自建
OURWORD_SITE=/path/to/your/checkout python3 ourword_mcp.py
Comments
New to MCP?
From what the protocol is to installing your first server, these answers cover most first questions.
An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.