OmniContext
Connects Claude Desktop to Notion and YouTube transcripts with one-click installation, enabling natural language queries about your pages and video content.
README
OmniContext
Connect Claude to your life — without touching a terminal.
<!-- INSERTA AQUÍ EL GIF DE DEMOSTRACIÓN -->
OmniContext is a local MCP server for Claude Desktop that connects Claude to the tools you already use — Notion and YouTube today, more on the way. Install it as a one-click extension, no terminal or API dashboard required.
You: Summarize this video for me: youtube.com/watch?v=xxxxxxxxxxx
Claude: [reads the transcript through OmniContext and replies instantly]
You: What does my Notion page "Q3 Roadmap" say about launch dates?
Claude: [searches your Notion, opens the page, answers from its content]
Features
- Notion, without the copy-pasting — search page titles and pull full page content as clean markdown, right from the conversation.
- YouTube videos, the moment you paste the link — Claude reads the transcript and can summarize, analyze, or answer questions about it instantly.
- Clean transcript exports — when you want the raw text (for notes, research, or another tool), OmniContext saves it as a tidy
.mdfile instead of dumping it into the chat.
More integrations are on the way — see the Roadmap.
Installation
No code, no terminal, no API dashboard.
- Download the latest
omnicontext.mcpbfrom the Releases tab. - Open Claude Desktop → the menu (☰) →
File → Settings → Extensions. - Click Advanced settings then scroll down to Install Extension… and select it.
- Using Notion? Paste your integration token into the form Claude Desktop shows you. That's the entire setup.
- Try it in a chat:
- "What's this YouTube video about:
<link>" - "Read my Notion page:
<link>"
- "What's this YouTube video about:
Note: your Notion token is stored encrypted by your OS (Keychain on macOS, Credential Manager on Windows) — never in a plain text file. If you don't need Notion, skip it; the YouTube tools work with zero setup.
Development
For running the server from source, contributing, or wiring it into Claude Code instead of Claude Desktop.
git clone <your-repo-url>
cd omnicontext
npm install
npm run build # compiles TypeScript -> dist/
npm test # full suite, incl. live smoke tests
npm run pack # builds omnicontext.mcpb locally
Register it with Claude Code:
claude mcp add --transport stdio omnicontext \
--env NOTION_TOKEN=your_notion_token \
-- node /absolute/path/to/omnicontext/dist/index.js
Run it directly during development:
npm run dev
Debug tool calls with the official MCP Inspector:
npx @modelcontextprotocol/inspector node dist/index.js
Tools reference
| Tool | What it does |
|---|---|
notion_get_page |
Reads a Notion page (URL or ID) and returns it as markdown |
notion_search |
Searches page titles across everything shared with the integration |
youtube_get_transcript |
Fetches a YouTube video's transcript as plain text, for in-chat analysis |
youtube_save_transcript |
Saves a video's full transcript to a local .md file |
Roadmap
OmniContext is an early MVP — the foundation is in place, and the integrations list is just getting started.
- [x] YouTube transcripts (in-chat + file export)
- [x] Notion (internal token, read-only)
- [x] One-click
.mcpbpackaging for Claude Desktop - [ ] Notion OAuth — one-click connect, no more manual tokens or sharing pages by hand
- [ ] Discord integration
- [ ] Telegram integration
- [ ] More integrations, one context source at a time
Privacy & security
OmniContext runs entirely on your machine. It talks directly to Notion's and YouTube's own APIs — there is no OmniContext backend, no analytics, and no middleman server that sees your data.
Contributing
Issues and pull requests are welcome. If you're adding a new integration, follow the existing pattern in src/integrations/<name>/ — each one is self-contained and registers its own tools without touching the core server.
Support
Found a bug or want to request an integration? Open an issue.
License
MIT — free to use, modify, and self-host.
Built on the Model Context Protocol.
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。