Intelligent YouTube video analysis with token-optimized summaries and reporting
π English | νκ΅μ΄
MCP YouTube Intelligence
YouTube μμμ μ§λ₯μ μΌλ‘ λΆμνλ MCP μλ² + CLI
MCP (Model Context Protocol)λ Claude, Cursor κ°μ AI λκ΅¬κ° μΈλΆ μλΉμ€λ₯Ό μ¬μ©ν μ μκ² ν΄μ£Όλ νμ€ νλ‘ν μ½μ λλ€. μ΄ μλ²λ₯Ό μ°κ²°νλ©΄ "μ΄ μμ μμ½ν΄μ€" νλ§λλ‘ λΆμμ΄ μλ£λ©λλ€.
π― ν΅μ¬ κ°μΉ: μλ³Έ μλ§(2,000~30,000 ν ν°)μ μλ²μμ μ²λ¦¬νμ¬ LLMμλ ~200β500 ν ν°λ§ μ λ¬ν©λλ€.
π€ μ μ΄ μλ²μΈκ°?
λλΆλΆμ YouTube MCP μλ²λ μλ³Έ μλ§μ κ·Έλλ‘ LLMμ λμ§λλ€.
| κΈ°λ₯ | κΈ°μ‘΄ MCP μλ² | MCP YouTube Intelligence |
|---|---|---|
| μλ§ μΆμΆ | β | β |
| μλ²μ¬μ΄λ μμ½ (ν ν° μ΅μ ν) | β | β |
| ꡬ쑰νλ 리ν¬νΈ (μμ½+ν ν½+μν°ν°+λκΈ) | β | β |
| μ±λ λͺ¨λν°λ§ (RSS) | β | β |
| λκΈ κ°μ± λΆμ | β | β |
| ν ν½ μΈκ·Έλ©ν μ΄μ | β | β |
| μν°ν° μΆμΆ (ν/μ 200+κ°) | β | β |
| μλ§/YouTube κ²μ | β | β |
| λ°°μΉ μ²λ¦¬ | β | β |
| SQLite/PostgreSQL μΊμ | β | β |
π λΉ λ₯Έ μμ
1. μ€μΉ
pip install mcp-youtube-intelligence
pip install yt-dlp # μλ§ μΆμΆμ νμ
π‘ LLM μμ΄λ κΈ°λ³Έ μμ½(ν΅μ¬ λ¬Έμ₯ μΆμΆ)μ λμν©λλ€. κ³ νμ§ μμ½μ μνλ©΄ μλ LLM μ€μ μ μ°Έκ³ νμΈμ.
2. 첫 λ²μ§Έ λͺ λ Ήμ΄ μ€ν
# 리ν¬νΈ μμ± β μμ½, ν ν½, μν°ν°, λκΈμ νλ²μ λΆμ (LLM μ°λνμ)
mcp-yt report "https://www.youtube.com/watch?v=LV6Juz0xcrY"
# μλ§ μμ½λ§
mcp-yt transcript "https://www.youtube.com/watch?v=LV6Juz0xcrY"
# μμ IDλ§ μ¨λ λ©λλ€
mcp-yt report LV6Juz0xcrY
β οΈ zsh μ¬μ©μ: URLμ
?κ° μμΌλ―λ‘ λ°λμ λ°μ΄νλ‘ κ°μΈμΈμ.
π 리ν¬νΈ μΆλ ₯ μμ
mcp-yt report "https://www.youtube.com/watch?v=LV6Juz0xcrY" μ€ν κ²°κ³Ό (extractive μμ½):
# πΉ Video Analysis Report: OpenClaw Use Cases that are Actually Helpful! (ClawdBot)
> Channel: Duncan Rogoff | AI Automation | Duration: 16:29 | Language: en_ytdlp
## 1. Summary
OpenClaw is the most powerful AI agent framework in the world right now and
it's about to replace your entire workflow. I spent over $200 in the last
48 hours stress testing the system so you don't have to. It defines who it
is, how it behaves, and crucial behavioral boundaries. If you think open
claw is cool, just check out this video up here of 63 insane use cases
that other people are doing.
## 2. Key Topics
| # | Topic | Keywords | Timespan |
|---|-------|----------|----------|
| 1 | framework, world, right | framework, world, right | 0:00~0:05 |
| 2 | like, really, there | like, really, there | 0:05~2:23 |
| 3 | like, max, using | like, max, using | 2:23~4:22 |
| 4 | going, like, something | going, like, something | 4:22~5:03 |
| 5 | like, agents, basically | like, agents, basically | 5:03~6:04 |
| ... | ... | ... | ... |
| 15 | think, open, claw | think, open, claw | 16:24~16:29 |
## 4. Keywords & Entities
- **Technology**: GitHub, LLM, GPT
- **Company**: Anthropic, Apple
## 5. Viewer Reactions
- Total comments: 20
- Sentiment: Positive 45% / Negative 0% / Neutral 55%
- Top opinions:
- **@geetee2583** (positive, π8): Great info. Just need your inset video out of the way...
- **@bdog4026** (positive, π3): This tool is wild! Definitely the most in depth explanation...
- **@magalyvilela4917** (neutral, π3): Came to this video wondering it gonna teach me how to set up...
π CLI μ 체 λͺ λ Ήμ΄
π 리ν¬νΈ (ν΅μ¬ κΈ°λ₯)
β οΈ **리ν¬νΈμ μμ½ μΉμ μ LLM μ°λμ΄ νμμ λλ€. Ollama λΉ λ₯Έ μ€μ (무λ£, 3λΆμ΄λ©΄ λ):
# 1. Ollama μ€μΉ: https://ollama.ai # 2. λͺ¨λΈ λ€μ΄λ‘λ ollama pull qwen2.5:7b # 3. νκ²½λ³μ μ€μ export MYI_LLM_PROVIDER=ollama export MYI_OLLAMA_MODEL=qwen2.5:7b # μ격 μλ²λΌλ©΄ νΈμ€νΈλ μ§μ export MYI_OLLAMA_BASE_URL=http://your-server:11434
mcp-yt report "https://youtube.com/watch?v=VIDEO_ID"
mcp-yt report VIDEO_ID --provider ollama # LLM νλ‘λ°μ΄λ μ§μ
mcp-yt report VIDEO_ID --no-comments # λκΈ μ μΈ
mcp-yt report VIDEO_ID -o report.md # νμΌ μ μ₯
π― μλ§ μΆμΆ + μμ½
mcp-yt transcript VIDEO_ID # μμ½ (~200β500 ν ν°)
mcp-yt transcript VIDEO_ID --mode full # μ 체 μλ§
mcp-yt transcript VIDEO_ID --mode chunks # μ²ν¬ λΆν
mcp-yt --json transcript VIDEO_ID # JSON μΆλ ₯
κΈ°ν
mcp-yt video VIDEO_ID # λ©νλ°μ΄ν°
mcp-yt comments VIDEO_ID --max 20 # λκΈ (κ°μ± λΆμ ν¬ν¨)
mcp-yt entities VIDEO_ID # μν°ν° μΆμΆ
mcp-yt segments VIDEO_ID # ν ν½ μΈκ·Έλ©ν
μ΄μ
mcp-yt search "ν€μλ" --max 5 # YouTube κ²μ
mcp-yt monitor subscribe @μ±λνΈλ€ # μ±λ λͺ¨λν°λ§
mcp-yt playlist PLAYLIST_ID # νλ μ΄λ¦¬μ€νΈ
mcp-yt batch ID1 ID2 ID3 # λ°°μΉ μ²λ¦¬
mcp-yt search-transcripts "ν€μλ" # μ μ₯λ μλ§ κ²μ
π‘ λͺ¨λ λͺ λ Ήμ΄μ
--jsonνλκ·Έλ₯Ό μΆκ°νλ©΄ JSON μΆλ ₯λ©λλ€.
π MCP μλ² μ°κ²°
MCP μλ²λ stdio νλ‘ν μ½λ‘ ν΅μ ν©λλ€.
Claude Desktop / Cursor / OpenCode
μ€μ νμΌμ μΆκ° (claude_desktop_config.json, .cursor/mcp.json, mcp.json):
{
"mcpServers": {
"yout
Tools (4)
reportGenerates a comprehensive video analysis report including summary, topics, entities, and comments.transcriptExtracts and summarizes video transcripts with token optimization.searchPerforms a YouTube search for videos based on keywords.monitorSubscribes to a channel for monitoring updates.Environment Variables
MYI_LLM_PROVIDERThe LLM provider to use for analysis (e.g., ollama)MYI_OLLAMA_MODELThe specific Ollama model to useMYI_OLLAMA_BASE_URLThe base URL for the Ollama serverConfiguration
{"mcpServers": {"youtube-intelligence": {"command": "mcp-yt", "args": ["stdio"]}}}