Awesome MCP › Other MCP Servers

lfnovo/content-core

⭐ 172 Jupyter Notebook repository created 2025-04-14

Content Core extracts, processes and summarises content from URLs, files and raw text, and offers the same functionality through an async Python API, a CLI and a Model Context Protocol server. Supported inputs cover web pages, YouTube videos and Reddit posts, documents in PDF, DOCX, PPTX, XLSX, EPUB, Markdown and plain text, and media files in MP3, WAV, M4A, FLAC and OGG for audio and MP4, AVI, MOV and MKV for video. The MCP server is started with content-core mcp, or without installing anything through uvx content-core mcp, and is registered in claude_desktop_config.json like any other stdio server. It exposes two tools, extract_content and summarize_content, both returning plain text, so an assistant can pull the readable content of an arbitrary link or document into the conversation and condense it. Behind the tools sits a pluggable engine layer: URL extraction can run in auto mode or use simple fetching, Firecrawl, Jina or Crawl4AI, and document parsing can use a simple reader or Docling. Summarisation and speech-to-text go through the Esperanto abstraction, so OpenAI, Anthropic, Google, Groq, DeepSeek, Ollama and other providers can be swapped by configuration alone. Settings are resolved from constructor arguments, then CCORE_ environment variables, then a TOML file in the home directory, then defaults, and standard proxy variables are honoured automatically. The repository also ships an Agent Skill describing how an agent should use the tool, and doubles as a plugin marketplace so the skill installs natively in Claude Code and Codex, with a manual copy of SKILL.md as a fallback. The project is MIT licensed and published on PyPI. It suits anyone who needs one consistent path from a mixed pile of links, documents and recordings to clean text.

https://github.com/lfnovo/content-core

content-extractionsummarizationmcp-serverpythonpdftranscriptionscraping