GitHub
Features Workflows Benchmarks Docs
v1.0.0 Released No YouTube API Keys Required

Teach your AI using
YouTube Videos

Lightweight, AI-first YouTube extraction CLI for transcripts, subtitles, chapters, descriptions, and structured metadata. Ready for RAG pipelines, MCP servers, and LLM coding agents.

npm i -g vidilearn
Get Started
0 API Keys Required
< 1.5s Extraction Time
100% AI-Ready JSON
bash - vidilearn extract
~ $ |

Engineered for AI Autonomy

A simple, lightweight CLI designed to bypass API restrictions and feed structured data straight to RAG databases, MCP servers, and LLM coding assistants.

No API Keys Required

Zero configuration. Bypasses complex GCP Project setups, YouTube OAuth verification, and query quota limits. Just install and run.

Structured JSON Output

Get transcripts, metadata, channel details, custom timestamps, and chapters in a single clean JSON payload optimized for vector storage.

Zero Browser Automation

No Puppeteer, Playwright, or Selenium required. Extremely low RAM and CPU footprint, meaning it will run smoothly inside serverless functions or containerized jobs.

Built for AI Workflows

Perfect fit for OpenAI Agents, MCP Server implementations, Codex CLI workflows, CrewAI, AutoGen, and LangGraph structures. Ingest whole channels or tutorials as technical contexts.

Local-First Processing

Extract and process content entirely on your machine without relying on external APIs or cloud processing pipelines. Full control over privacy, speed, and reliability.

Production CLI Experience

Designed for modern developer workflows with clean commands, structured outputs, automation-ready pipelines, and seamless integration into CI/CD systems.

Flexible CLI Integration

From simple metadata extraction to automated developer workflows, customize Vidilearn to feed your pipeline needs.

Basic Extraction Examples 
# Standard complete extraction (prints title, transcript, chapters, etc.)
vidilearn extract "https://youtube.com/watch?v=VIDEO_ID"

# Pretty-print formatted JSON
vidilearn extract "https://youtube.com/watch?v=VIDEO_ID" --pretty

# Save structured JSON to file directly
vidilearn extract "https://youtube.com/watch?v=VIDEO_ID" > context.json

# Specialized extractions
vidilearn extract "https://youtube.com/watch?v=VIDEO_ID" --transcript
vidilearn extract "https://youtube.com/watch?v=VIDEO_ID" --chapters
vidilearn extract "https://youtube.com/watch?v=VIDEO_ID" --metadata
RULE OF THUMB

Always wrap URLs in double quotes (e.g. "https://youtube.com/...") to prevent terminal shells from parsing special characters like & or ? as background job triggers.

Example JSON Schema Response 
{
  "title": "Build AI Agents",
  "channel": "AI Academy",
  "duration": "12:45",
  "description": "Learn how to build and orchestrate multi-agent architectures...",
  "transcript": "In this video, we will design and code multi-agent LLM teams...",
  "chapters": [
    {
      "title": "Introduction",
      "timestamp": "00:00"
    },
    {
      "title": "Agent Architecture Design",
      "timestamp": "03:42"
    },
    {
      "title": "Code Implementation",
      "timestamp": "07:15"
    }
  ]
}
MCP Integration & Pipelines 
# Feed structured data straight into local vector database loading scripts
vidilearn extract "https://youtube.com/watch?v=VIDEO_ID" --json | node ingest-vector-store.js

# Use directly inside MCP (Model Context Protocol) tool configurations:
{
  "mcpServers": {
    "vidilearn-server": {
      "command": "npx",
      "args": ["-y", "vidilearn", "extract", "{{url}}", "--json"]
    }
  }
}
Contextual Learning for Coding Agents 
# Extract technical documentation from standard video tutorials
vidilearn extract "https://youtube.com/watch?v=VIDEO_ID" > /tmp/context-learning.json

# Feed the output as context injection to Codex or Gemini commands
codex run "implement a backend auth system using the rules in context-learning.json" \
  --context=/tmp/context-learning.json

High-Performance Statistics

Vidilearn is designed to optimize resources. See how we rank against traditional web scrapers and API integrations.

Extraction Speed (s)

Lower is better (seconds per extraction)

Browser Automation (Puppeteer) 7.2s
Official YouTube Data API v3 3.5s
Vidilearn Engine 1.2s

Startup Overhead / Install Size

Lower is better (Disk usage)

Traditional Scrapers (Chromium) 312 MB
Google Cloud SDK Wrapper 48 MB
Vidilearn Package 4.2 MB

Command Reference

Quick look up of all available flags, arguments, and roadmap features.

Command / Flag Description Default Behavior
vidilearn extract <url> Core parsing command for a YouTube video. Extracts all available data
--pretty Formats JSON output with spaces for human readability. Minified JSON string
--transcript Extracts only the transcription text. Extracts full dataset
--chapters Returns video chapters & timestamps. Extracts full dataset
--metadata Returns core title, description, channel, and duration. Extracts full dataset
--help Prints standard usage, examples, and flags information. -

Roadmap & Future Modules

  • Streaming transcript extraction (Real-time console printing)
  • Interactive MCP server compatibility layer
  • Batch playlist processing & full channel extraction
  • Local embedding generation pipelines (Ready-to-query chunks)
  • Multi-language subtitle translations and support
  • Vector Database upload integrations (Pinecone, Chroma, pgvector)