fetchtranscript-mcp
Official# FetchTranscript MCP Server
<!-- mcp-name: com.fetchtranscript/fetchtranscript -->
An [MCP](https://modelcontextprotocol.io) server that gives AI assistants (Claude Desktop, Cursor, etc.) access to YouTube transcripts, video metadata, channels, and search via the [FetchTranscript](https://fetchtranscript.com) API.
## Features
**8 Tools** — get transcripts, list languages, video metadata, chapters, comments, channel info, channel videos, search
**4 Documentation Resources** — API reference, quickstart guide, pricing, workflow examples
**3 Prompt Templates** — summarize video, compare videos, extract quotes
## Quick Start
### 1. Get an API Key
Sign up at [fetchtranscript.com](https://fetchtranscript.com) and create an API key from the dashboard.
### 2. Configure Claude Desktop
Add to your `claude_desktop_config.json`:
```json
{
"mcpServers": {
"fetchtranscript": {
"command": "uvx",
"args": ["fetchtranscript-mcp"],
"env": {
"FETCHTRANSCRIPT_API_KEY": "yt_your_key_here"
}
}
}
}
```
### 3. Use it
Ask Claude:
- "Get the transcript of YouTube video dQw4w9WgXcQ"
- "Summarize this YouTube video: https://youtube.com/watch?v=..."
- "Search YouTube for 'machine learning explained' and summarize the top result"
## Configuration for Other Clients
### Cursor
Add to `.cursor/mcp.json`:
```json
{
"mcpServers": {
"fetchtranscript": {
"command": "uvx",
"args": ["fetchtranscript-mcp"],
"env": {
"FETCHTRANSCRIPT_API_KEY": "yt_your_key_here"
}
}
}
}
```
### Claude Code
```bash
claude mcp add fetchtranscript -- uvx fetchtranscript-mcp
```
Set `FETCHTRANSCRIPT_API_KEY` in your environment.
## Installation Alternatives
### pip
```bash
pip install fetchtranscript-mcp
FETCHTRANSCRIPT_API_KEY=yt_xxx fetchtranscript-mcp
```
### npx
```bash
FETCHTRANSCRIPT_API_KEY=yt_xxx npx fetchtranscript-mcp
```
## Tools
| Tool | Description | Key defaults |
|------|-------------|--------------|
| `get_transcript` | Fetch video transcript | format=text |
| `list_languages` | Available transcript languages | — |
| `get_video_metadata` | Video title, views, likes, etc. | — |
| `get_chapters` | Video chapter list | — |
| `get_comments` | Top comments | limit=20 |
| `get_channel_info` | Channel metadata | — |
| `list_channel_videos` | Videos from a channel | limit=10 |
| `search_videos` | Search YouTube | limit=5 |
## Environment Variables
| Variable | Required | Description |
|----------|----------|-------------|
| `FETCHTRANSCRIPT_API_KEY` | Yes | Your API key (starts with `yt_`) |
| `FETCHTRANSCRIPT_BASE_URL` | No | Override API base URL (default: `https://api.fetchtranscript.com`) |
## Development
```bash
cd mcp
pip install -e ".[dev]"
pytest tests/
```
Test with MCP Inspector:
```bash
FETCHTRANSCRIPT_API_KEY=yt_xxx mcp dev src/fetchtranscript_mcp/server.py
```
## License
MIT
TDQS
Scored across 8 tools
Each tool targets a clearly distinct resource or action: transcript, comments, metadata, chapters, languages, channel info, channel video listing, and search. There is no meaningful overlap between any pair, and argument shapes reinforce the distinctions.
All tool names use a consistent snake_case verb_noun pattern (get_, list_, search_) with predictable resource suffixes. The use of get_ vs list_ vs search_ is semantically coherent and easy to follow.
Eight tools is well-scoped for a YouTube transcript and metadata server, with each tool earning its place. Nothing feels redundant or conspicuously absent from the count perspective.
The surface covers the core lifecycle well: transcript retrieval, language discovery, comments, metadata, chapters, channel info, channel videos, and search. Minor gaps exist—e.g., no comment-reply retrieval or in-transcript search—but agents can work around these for most transcript-focused workflows.