Rippr: a YouTube transcript ripper built for AI workflows
A Chrome extension that extracts transcripts from any YouTube video and saves them as clean files for RAG pipelines, LLM workflows, or just reading. Paste a URL, pick a format, download. Also summarizes videos with AI.
- Type
- Personal project
- Stack
- Claude Code + Chrome Manifest V3
- Formats
- RAG (.txt), Structured (.json), Readable (.md)
Why I Built This
I was watching a lot of YouTube videos for research, mostly about job hunting in Japan, AI trends, and design career strategies. I kept needing the transcripts so I could feed them into Claude or ChatGPT for analysis. But extracting transcripts manually was painful, and every existing tool I found was either covered in ads or gave me messy output that needed cleanup before I could use it. So I built my own. Something that gives me clean text in one click, in formats that are actually designed for AI workflows.
Rippr workflow
Design Decisions
01
Three output formats for three use cases
RAG mode outputs a single continuous text block, optimized for chunking and embedding. Structured mode gives you timestamped JSON with metadata. Readable mode exports clean markdown. Each one is designed for a specific workflow instead of just being a generic "download transcript" button.
02
AI summarization built in
Instead of just ripping the transcript, I added a Summarize tab that processes the full transcript through AI and returns a title, summary, and key bullet points. For a 2-hour podcast, this turns a wall of text into something you can scan in 30 seconds.
03
Runs entirely in the browser
No data gets sent to any external server. Transcripts are extracted and downloaded locally. The only network request is to YouTube's own APIs. No accounts, no tracking, no analytics. I wanted something I'd actually trust using myself.
04
MCP server for AI agents
Beyond the extension and website, Rippr is also available as an MCP (Model Context Protocol) server. Any AI agent, whether it's Claude, Cursor, or a custom build, can call npx rippr-mcp to extract transcripts programmatically. This turns Rippr from a tool I use manually into infrastructure that AI workflows can plug into directly.
The MCP server was a deliberate experiment in building for a different kind of user. I wanted to explore what it takes to create interfaces that AI agents consume, not just humans.