# YouTube Intelligence

[![Listed on mcpservers.org](https://mcpservers.org/badge.svg)](https://mcpservers.org/servers/trichanizar/youtube-intelligence.git)

**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.

![demo](demo.gif)

---

## 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

```bash
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

```bash
cp .env.example .env
```

Open `.env` and add your free YouTube Data API v3 key.  
→ Get one at [console.cloud.google.com](https://console.cloud.google.com) (free, takes 2 min)  
→ Full walkthrough: [SETUP.md](SETUP.md)

### 3. Connect to Claude Desktop

Add this to your Claude Desktop config (`~/Library/Application Support/Claude/claude_desktop_config.json`):

```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)

```bash
# 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.

---

## Roadmap

- [ ] Web interface
- [ ] Scheduled re-indexing (new videos auto-indexed)
- [ ] Export insights to Notion / PDF
- [ ] Multi-language synthesis