search-youtubeLABSCO SUMMARY
It builds a searchable library out of the videos you have already watched.
Fetching one transcript is a small thing; the difference here is that the transcripts stay. Once a dozen talks are stored, the question becomes "which of these mentioned the migration" rather than "which video was that in".
What it is
A Python server that reaches YouTube without the official API. It searches videos, retrieves transcripts, and stores video information and transcript text in a local vector database so later questions run as semantic search over what you have already collected.
What you get
- YouTube search that does not go through the official Data API
- A video's transcript by URL
- Video information and its transcript stored in a vector database on your machine
- Semantic search across the transcripts you have stored, by meaning rather than keyword
Requirements
Python 3.10 or newer (the project's pyproject.toml requires it, although the README says 3.8), the uv package manager, and a Google API key in the environment — it is used for the embeddings behind the semantic search, not for YouTube access.
Setup effort
Build from source — git clone https://github.com/IA-Programming/Youtube-MCP.git
USE CASES
01
Finding the video that covers something
Search YouTube from the conversation and get candidates back without opening the site.
02
Reading a talk instead of watching it
Pull the transcript by URL and work from the text.
03
Building a reference library from a series
Store each video's information and transcript as you go, so the collection grows into something searchable.
04
Answering a question across everything you have collected
Run a semantic search over the stored transcripts and find the passage even when the wording is different.
4 tools
Listed in project docs. Your client may show a different list.
get-transcriptstore-video-infosearch-transcriptsBEFORE YOU INSTALL
What to know first
Tools you get
4 tools named in the project's documentation.
Requires
- uv on PATH
- Python 3.10+
- GOOGLE_API_KEY in your environment
Where it runs
Both documented setups run the server on your machine.
INSTALLATION
Connect your client
Server setup
TERMINAL
git clone https://github.com/IA-Programming/Youtube-MCP.git
cd Youtube-MCP
uv syncMCP Clients
Claude Desktop — uv
CLAUDE DESKTOP — UV · JSON
{
"mcpServers": {
"youtube": {
"command": "uv",
"args": [
"--directory",
"<YOUR_PATH_TO_PROJECT>",
"run",
"server.py"
],
"env": {
"GOOGLE_API_KEY": "<YOUR_GOOGLE_API_KEY>"
}
}
}
}<YOUR_PATH_TO_PROJECT> is the full path of the folder the clone created; on macOS or Linux that is a path such as /Users/you/Youtube-MCP.
Claude Desktop — venv Python
CLAUDE DESKTOP — VENV PYTHON · JSON
{
"mcpServers": {
"youtube": {
"command": "<YOUR_PATH_TO_PROJECT>\\.venv\\Scripts\\python.exe",
"args": [
"<YOUR_PATH_TO_PROJECT>\\server.py"
],
"env": {
"GOOGLE_API_KEY": "<YOUR_GOOGLE_API_KEY>"
}
}
}
}This is the Windows layout the README gives. <YOUR_PATH_TO_PROJECT> is the folder the clone created. On macOS or Linux the interpreter is .venv/bin/python, so use the full paths of .venv/bin/python and server.py inside your clone.
