YouTube Research MCP
Turn YouTube videos into citation-ready research for Codex, Claude, Cursor, and other MCP clients. Paste a video URL and get transcript evidence with timestamps and links that open at the exact quoted moment.```text
Input: https://youtu.be/dQw4w9WgXcQ
Output: [01:05] ...evidence text...
https://www.youtube.com/watch?v=dQw4w9WgXcQ&t=65s
No YouTube API key is required for transcript research.
## Why this server
- **Citation-ready research** — every transcript segment includes a timestamp and source URL.
- **Zero-key quick start** — transcripts work immediately; add an API key only for search and analytics.
- **URL-first input** — accepts normal, short, Shorts, embed, live, and raw video-ID formats.
- **Agent-efficient filtering** — search inside transcripts and return nearby context instead of spending tokens on the entire video.
- **Full YouTube intelligence** — optional API mode adds comments, video/channel statistics, trends, and comparisons.
- **Remote-native MCP** — Streamable HTTP transport at `/mcp`, plus Docker and Smithery support.
## Architecture and trust boundary```mermaid
flowchart LR
C[MCP client] -->|stdio or Streamable HTTP| T[Transport]
T --> R[Tool and resource registry]
R --> S[YouTubeService]
S --> K[(In-memory cache)]
S -->|no key required| YT[Caption tracks]
S -->|optional API key| API[YouTube Data API v3]
The MCP layer validates inputs and returns structured evidence; it does not ask a model to invent summaries or citations. Transcript tools derive timestamps and canonical links from caption segments. Data API tools return an explicit key-required error when YOUTUBE_API_KEY is unavailable, and remote authentication is enforced before MCP session handling.
Focused evidence, not a transcript dump
The included live smoke test asks a focused question about a public video and compares the response with the full timestamped transcript:
| Response | Characters returned |
|---|---|
| Full transcript | 85,518 |
research-video (3 citations with source identity) |
1,550 |
| Reduction | 98.2% |
This measures response characters, not model-specific tokens. Reproduce it against the default public fixture—or substitute your own video and query:```bash
npm run test:live
LIVE_TEST_VIDEO='https://youtu.be/VIDEO_ID' LIVE_TEST_QUERY='evaluation' npm run test:live
## Quick start — no API key
Requirements: Node.js 22 or newer.
The shortest local install uses stdio and needs no API key:```bash
codex mcp add youtube-research -- npx -y @coyasong/youtube-mcp-server@latest
For Claude Desktop and other MCPB-compatible desktop clients, download the
latest one-click MCP bundle
and open it. The bundle vendors its runtime dependencies, starts locally over
stdio, and asks for a YouTube API key only if you want the optional analytics
tools.
Or install from the VS Code command line:```bash
code --add-mcp '{"name":"youtube-research","command":"npx","args":["-y","@coyasong/youtube-mcp-server@latest"]}'
For Claude Desktop, Cursor, and other stdio clients, use command `npx` with arguments `-y @coyasong/youtube-mcp-server@latest`.
To run the Streamable HTTP transport instead:```bash
git clone https://github.com/coyaSONG/youtube-mcp-server.git
cd youtube-mcp-server
npm install
npm run build
npm start
The server starts at http://localhost:3000/mcp in transcript-only mode. Confirm it with:```bash
curl http://localhost:3000/health
### Connect from Codex
With the HTTP server running:```bash
codex mcp add youtube-research --url http://localhost:3000/mcp
Then ask Codex:```text
Use research-video to find what this video says about evaluation,
and cite the exact moments: https://www.youtube.com/watch?v=VIDEO_ID
### Connect with MCP Inspector```bash
npx @modelcontextprotocol/inspector
# Connect to http://localhost:3000/mcp
Install through Smithery```bash
npx -y smithery@latest auth login
npx -y smithery@latest mcp add coyaSONG/youtube-mcp-server --client claude
## Enable search, comments, and analytics
Set a YouTube Data API v3 key to switch from `transcript-only` to `full` mode:```bash
export YOUTUBE_API_KEY=your_key_here
npm start
YOUTUBE_API_KEY is optional. PORT defaults to 3000. For stdio clients, set YOUTUBE_API_KEY in the MCP server environment configuration.
Secure a remote deployment
Do not expose a full-mode server publicly without authentication: unauthenticated users could consume your YouTube API quota. Set a strong bearer token and restrict browser origins when deploying outside localhost:```bash
export MCP_BEARER_TOKEN='replace-with-a-long-random-secret'
export CORS_ORIGIN='https://your-client.example'
export MAX_SESSIONS=100
export SESSION_IDLE_TIMEOUT_MS=1800000
npm start
Browser origins are denied by default. `CORS_ORIGIN` accepts a comma-separated
allowlist of exact `http://` or `https://` origins; wildcard origins are rejected.
Connect Codex using an environment variable rather than writing the secret into its configuration:```bash
export YOUTUBE_MCP_TOKEN='replace-with-a-long-random-secret'
codex mcp add youtube-research \
--url https://your-server.example/mcp \
--bearer-token-env-var YOUTUBE_MCP_TOKEN
/health remains public for container health checks. MCP requests return 401 when authentication is enabled and the bearer token is missing or invalid. Idle sessions are removed automatically, and MAX_SESSIONS bounds memory use.
Best first tool
research-video accepts:
video: a YouTube URL or 11-character video IDlanguage: optional caption language such asen,ko, orjaquery: optional phrase to find inside the transcriptcontextLines: surrounding segments to retain, from 0 to 5matchMode:word(default) orsubstringstartSeconds/endSeconds: optional time windowoffset: result offset for paginationmaxSegments: response cap from 1 to 1,000 (default: 200)
It returns structured JSON containing the video title and channel identity, canonical video URL, full caption-track duration and segment count, matching transcript segments, human-readable citation labels, timestamps, directly navigable citation URLs, and pagination metadata. For long videos, use a query or time window first; follow nextOffset only when more evidence is needed.
Compare evidence across videos
research-videos applies one focused query to 2–5 video URLs concurrently. It returns the same structured, timestamp-linked evidence for each source while capping results per video. This is useful for comparing interviews, checking whether multiple sources support a claim, or researching a topic across a short watchlist.```json
{
"videos": [
"https://youtu.be/VIDEO_ONE",
"https://youtu.be/VIDEO_TWO"
],
"query": "evaluation",
"maxSegmentsPerVideo": 10
}
See YouTube Research Recipes for copy-paste workflows for
fact-checking claims, comparing interviews, navigating long talks, researching
multilingual captions, and building citation-ready notes.
## Capability modes
| Capability | No-key mode | With `YOUTUBE_API_KEY` |
|---|:---:|:---:|
| Transcript research, filtering, key moments, segmentation, summaries | Yes | Yes |
| Video search and comments | No | Yes |
| Video/channel metadata, statistics, trends, and comparisons | No | Yes |
Captions must be available for the requested video. Age-restricted, private, region-restricted, or caption-disabled videos may not return a transcript.
If this project saves you research time, consider
starring the repository so
other agent builders can discover it.
See the broader AI tooling portfolio for the
reliability principles shared with `ralph-research` and `tmuxicate`.
## Docker```bash
docker build -t youtube-research-mcp .
docker run --rm -p 3000:3000 youtube-research-mcp
# Full mode
docker run --rm -p 3000:3000 -e YOUTUBE_API_KEY=your_key youtube-research-mcp
Development```bash
npm run dev # HTTP server from TypeScript
npm test # build and run all tests
npm run test:live # live public-video transcript and citation smoke test
npm run test:user # clean-room smoke test against the published npm package
npm run test:mcpb # build, unpack, and exercise the installable MCP bundle
Maintainers can follow the release guide. Published GitHub
releases run the complete npm, MCP Registry, MCPB, and Smithery delivery pipeline.
## API Reference
### Resources
- `youtube://video/{videoId}` - Get detailed information about a specific video
- `youtube://channel/{channelId}` - Get information about a specific channel
- `youtube://transcript/{videoId}` - Get transcript for a specific video
- Optional query parameter: `?language=LANGUAGE_CODE` (e.g., `en`, `ko`, `ja`)
### Tools
#### Basic Tools
- `research-video` - Get citation-ready transcript evidence from a URL or video ID without an API key
- `research-videos` - Compare timestamp-linked evidence across 2–5 videos without an API key
- `search-videos` - Search for YouTube videos with advanced filtering options
- `get-video-comments` - Get comments for a specific video
- `get-video-transcript` - Get transcript for a specific video with optional language
- `enhanced-transcript` - Advanced transcript extraction with filtering, search, and multi-video capabilities
- `get-key-moments` - Extract key moments with timestamps from a video transcript for easier navigation
- `get-segmented-transcript` - Divide a video transcript into segments for easier analysis
Tools requiring `YOUTUBE_API_KEY` are search, comments, statistics, discovery, and channel analysis. `enhanced-transcript` needs the key only when `includeMetadata` is `true`.
#### Statistical Tools
- `get-video-stats` - Get statistical information for a specific video
- `get-channel-stats` - Get subscriber count, view count, and other channel statistics
- `compare-videos` - Compare statistics across multiple videos
#### Discovery Tools
- `get-trending-videos` - Retrieve trending videos by region and category
- `get-video-categories` - Get available video categories for a specific region
#### Analysis Tools
- `analyze-channel-videos` - Analyze performance trends of videos from a specific channel
### Prompts
- `video-analysis` - Generate an analysis of a YouTube video
- `transcript-summary` - Generate a summary of a video based on its transcript with customizable length and keywords extraction
- `segment-by-segment-analysis` - Provide detailed breakdown of content by analyzing each segment of the video
## Examples
### Accessing a Video Transcript```
youtube://transcript/dQw4w9WgXcQ
Getting a Transcript in a Specific Language```
youtube://transcript/dQw4w9WgXcQ?language=en
### Using the Statistical Tools```javascript
// Get video statistics
{
"type": "tool",
"name": "get-video-stats",
"parameters": {
"videoId": "dQw4w9WgXcQ"
}
}
// Compare multiple videos
{
"type": "tool",
"name": "compare-videos",
"parameters": {
"videoIds": ["dQw4w9WgXcQ", "9bZkp7q19f0"]
}
}
Using the Transcript Summary Prompt```javascript
{
"type": "prompt",
"name": "transcript-summary",
"parameters": {
"videoId": "dQw4w9WgXcQ",
"language": "en"
}
}
### Using the Enhanced Transcript Tool```javascript
// Basic multi-video transcript extraction
{
"type": "tool",
"name": "enhanced-transcript",
"parameters": {
"videoIds": ["dQw4w9WgXcQ", "9bZkp7q19f0"],
"format": "timestamped"
}
}
// With search and time filtering
{
"type": "tool",
"name": "enhanced-transcript",
"parameters": {
"videoIds": ["dQw4w9WgXcQ"],
"filters": {
"timeRange": {
"start": 60, // Start at 60 seconds
"end": 180 // End at 180 seconds
},
"search": {
"query": "never gonna",
"contextLines": 2
}
},
"format": "merged"
}
}
// With duration-based segmentation for easier analysis
{
"type": "tool",
"name": "enhanced-transcript",
"parameters": {
"videoIds": ["dQw4w9WgXcQ"],
"filters": {
"segment": {
"count": 5,
"method": "smart" // Balances caption duration across segments
}
},
"format": "timestamped",
"language": "en"
}
}
Using the Enhanced Transcript Analysis Features```javascript
// Get key moments from a video
{
"type": "tool",
"name": "get-key-moments",
"parameters": {
"videoId": "dQw4w9WgXcQ",
"maxMoments": 5
}
}
// Get a segmented transcript
{
"type": "tool",
"name": "get-segmented-transcript",
"parameters": {
"videoId": "dQw4w9WgXcQ",
"segmentCount": 4
}
}
// Get a segment-by-segment analysis
{
"type": "prompt",
"name": "segment-by-segment-analysis",
"parameters": {
"videoId": "dQw4w9WgXcQ",
"segmentCount": 4
}
}
// Get customized transcript summary
{
"type": "prompt",
"name": "transcript-summary",
"parameters": {
"videoId": "dQw4w9WgXcQ",
"language": "en",
"summaryLength": "detailed",
"includeKeywords": true
}
}
```
Error Handling
The server handles various error conditions, including:
- Invalid or missing API key for Data API tools
- Video or channel not found
- Transcript not available
- Network issues
License
MIT
Acknowledgements
- Model Context Protocol TypeScript SDK
- YouTube Data API
- YouTube.js
Comment pagination and evidence limits
search-videos accepts the provider pageToken. get-video-comments preserves the next top-level token and accepts searchTerms for YouTube-side filtering.
Use includeReplies: true for the embedded reply subset, or completeReplies: true for bounded reply pagination. Full retrieval defaults to 100 replies per thread; maxRepliesPerThread accepts 1–1000. Across all threads, each call returns at most 1000 replies and 1 MiB of reply JSON and makes at most 20 additional reply requests. Each Data API request times out after 10 seconds and does not retry automatically.
Inspect each thread's replyCoverage: available is the provider's reported count, returned counts unique replies, complete confirms full observed coverage, and truncated indicates a local budget or pagination guard stopped collection. Counts may change while YouTube comments are being edited. Missing IDs or early provider termination do not establish complete coverage. Reduce the top-level maxResults to reserve more of the shared budget for each thread; pageToken continues top-level threads, not an incomplete reply list.
Research source metadata includes captionRequestedLanguage, captionTrackLanguage, captionLanguage, captionOrigin (manual, youtube_asr, or translated) and captionSourceOrigin. Unknown provenance is null.
