jztan/pdf-mcp🔥 active
▲ +3 ★ this week
Agentic RAG over one PDF or a whole folder, with page-level evidence to check every answer: hybrid search, selective page reads, tables, OCR, charts, and CJK layouts.
Universal LaTeX document skill for Claude Code: 27 templates, 27 scripts, 26 reference guides. Made with Claude Code on ✦ HappyCapy AI ✦ platform
$ git clone https://github.com/ndpvt-web/latex-document-skill.git▲ +3 ★ this week
Agentic RAG over one PDF or a whole folder, with page-level evidence to check every answer: hybrid search, selective page reads, tables, OCR, charts, and CJK layouts.
▲ +25 ★ this week
Let Claude (or any LLM) actually watch a video — scene-aware, deduplicated frames + transcript, from a URL or local file. Runs locally, MIT.
▲ +7 ★ this week
Self-hosted URL- and file-to-Markdown service for humans and AI agents - web pages, documents, images, audio, YouTube. PWA + REST + MCP + Claude Code skill, Reddit-aware, refreshable share links.
MCP server for computer use & browser automation - screenshot, OCR, click, type, find_text, Chrome/Electron CDP, template matching. macOS, Windows & Android. Works with Claude, Cursor, and any MCP client.
AI-Native document parser: PDF, Office & images → clean Markdown with LaTeX, tables & OCR. Zero-dependency CLI & skill for Claude Code, Cursor & AI agents.
MCP server that turns any video — YouTube, Instagram, TikTok, Loom, X, Vimeo, direct URLs, local files — into transcripts, key frames, OCR text, and metadata for AI agents.
Data from GitHub · snapshot Nov 16, 2026