Youtube-Intelligence
Access Youtube Intelligence Data from your favourite Youtubeur
Documentation
YouTube Intelligence
Not a transcript extractor. A knowledge access layer.
The creators you trust have the answers you need. Finding them takes hours. Now it takes seconds.
The real problem
You have a specific decision to make.
A meeting tomorrow. A deal to close. A hire to evaluate. A pitch to prepare.
The best answer probably exists — somewhere in the 500 hours of content
from the creators you already follow and trust.
But YouTube has no memory. No search within a channel. No synthesis across videos.
So you watch what comes up, miss 90% of what was said, and decide on incomplete information.
The knowledge is there. The access isn't.
What this does
You → "What does YC say about raising your seed round without traction?"
AI → Based on 3 videos:
1. "How to Get Your First Users" — Y Combinator (Mar 2023)
→ https://youtu.be/abc?t=312
"Most founders wait until the product is perfect before talking
to users. That's the wrong order. Launch ugly, learn fast..."
2. "Fundraising Fundamentals" — YC (Jan 2022)
→ https://youtu.be/xyz?t=89
"Investors fund lines, not dots. If you have no traction,
your job is to give them a second dot as fast as possible..."
One question. Instant answer. Real quotes. Exact timestamps.
What makes this different
| YouTube scraper tools | YouTube Intelligence | |
|---|---|---|
| Scope | Single video URL | Entire channel (1,000+ videos) |
| Search | None | Semantic — finds meaning, not keywords |
| Output | Raw transcript | Sourced answer with citations |
| Memory | None | Persistent — index once, query forever |
| Multi-channel | No | Yes — compare experts side by side |
| Your data | Cloud | 100% local — stays on your machine |
| Cost | Variable | Free (only a free YouTube API key) |
Use cases
- Founders — Extract everything a creator says about fundraising, GTM, hiring
- Sales reps — Mine 1,000 sales episodes for tactics on specific objections
- Researchers — Synthesize what multiple experts say about the same topic
- Executives — Stay on top of a creator's thinking without watching every video
Setup (5 minutes)
1. Clone & install
git clone https://github.com/trichanizar/youtube-intelligence
cd yt-intelligence
python -m venv .venv && source .venv/bin/activate
pip install -r requirements.txt
2. Add your YouTube API key
cp .env.example .env
Open .env and add your free YouTube Data API v3 key.
→ Get one at console.cloud.google.com (free, takes 2 min)
→ Full walkthrough: SETUP.md
3. Connect to Claude Desktop
Add this to your Claude Desktop config (~/Library/Application Support/Claude/claude_desktop_config.json):
{
"mcpServers": {
"yt-intelligence": {
"command": "/absolute/path/to/yt-intelligence/.venv/bin/python",
"args": ["/absolute/path/to/yt-intelligence/mcp_server.py"]
}
}
}
Restart Claude Desktop. Done.
Usage
Open Claude Desktop and start talking:
"Index the channel @GuillaumeMoubeche-FR"
→ Claude indexes all 68 videos (takes ~5 min)
"What does he say about the first 10 customers?"
→ Claude searches across all videos and answers with sources
"Now also index @30MPC and compare their views on cold outreach"
→ Claude searches both channels and synthesizes
No code. No terminal. Just conversation.
CLI usage (optional)
# Index a channel
python index_channel.py @GuillaumeMoubeche-FR
# Index with topic filter (skips off-topic videos)
python index_channel.py @30MPC --filter "B2B sales, prospecting, objection handling"
# Query directly
python search.py "how to handle pricing objections"
How it works
YouTube Channel
↓ YouTube Data API (free)
List of videos
↓ youtube-transcript-api (free)
Transcripts with timestamps
↓ Chunking + local embeddings (sentence-transformers, free)
Vector database (ChromaDB, local)
↓ Semantic search + cross-encoder reranking
Relevant excerpts
↓ Claude synthesizes
Sourced answer with timestamps
Everything runs locally. No data leaves your machine.
Requirements
- Python 3.10+
- Claude Desktop (free or Pro)
- YouTube Data API v3 key (free)
That's it.
