yt-mcp-agent

A YouTube video agent built using a custom MCP server and OpenAI's Agent's SDK. It can extract video transcripts given link fetch specialized instructions for things like: writing blog posts, video chapters, and social posts.

Resources:

• Talk recording

• Slides

This example is a prelude to Cohort 7 of the AI Builders Bootcamp.

Requirements

• Python 3.13+

• OpenAI API key

• uv package manager (recommended)

Related MCP server: YouTube Insights MCP Server

How to run this example

uv (recommended)

1. Clone the repository

   git clone <repository-url>
   cd yt-mcp-agent

2. Install dependencies with uv

   uv sync

3. Set up your OpenAI API key

   Create a .env file in the root directory:

   echo "OPENAI_API_KEY=your_api_key_here" > .env

4. Run the agent

   uv run main.py

5. Interact with agent

   Once running, you can ask the agent to analyze YouTube videos. Try prompts like:

   • "Summarize this: https://youtu.be/N3vHJcHBS-w?si=aw8PV0acYHJGPy7R"

   • "Generate chapter timestamps with links"

   • "Write me a LinkedIn post about the video"

Base Python/pip

1. Clone the repository

   git clone <repository-url>
   cd yt-mcp-agent

2. Create a virtual environment

   python -m venv venv
   source venv/bin/activate  # On Windows: venv\Scripts\activate

3. Install dependencies

   pip install -e .

4. Set up your OpenAI API key

   Create a .env file in the root directory:

   echo "OPENAI_API_KEY=your_api_key_here" > .env

5. Run the agent

   python main.py

6. Interact with agent