YouTube search, video data and transcripts for AI agents
YouTube MCP Server connects Claude, Cursor, Windsurf, and other MCP clients to public YouTube search results, video metadata, channel data, and timestamped transcripts. Hosted by HasData, it provides four read-only tools over Streamable HTTP. A HasData account replaces the need to configure a Google Cloud project or obtain a YouTube Data API key. Server details.
Use it for video research, competitor content analysis, and preparing summaries grounded in a video's transcript. The server retrieves structured data; your AI client writes the summaries, compares topics, and draws conclusions.
What can I ask it to do?
- "Find recent YouTube tutorials about web scraping, sorted by upload date and under four minutes long. Return titles, video IDs, and view counts."
- "Read the videos tab for @GoogleDevelopers. Group the returned upload titles by topic and link each example to its video."
- "Get the transcript for this video ID. Summarize the main argument with timestamps, and identify the language track used."
These prompts are example workflows. The agent may make several calls, particularly when retrieving additional pages or transcripts for multiple videos.
Four YouTube tools
| Tool | Input and result |
|---|
hasdata_youtube_search_getYoutubeSearchResults | Requires q. Search by keyword with optional sortBy, date, length, and videoType; use paginationToken for another page. Returns search results grouped by content type. Search parameters. |
hasdata_youtube_video_getYoutubeVideo | Requires v, the 11-character video ID. Retrieves video metadata, including title, description, engagement counts, and related videos where available. Video parameters. |
hasdata_youtube_channel_getYoutubeChannel | Requires channelId, accepting a channel ID or @handle. Select a tab, such as videos, shorts, or about; paginate when needed. Channel parameters. |
hasdata_youtube_transcript_getYoutubeTranscript | Requires v. Retrieves transcript segments and available language tracks. Use languageCode for an existing track and type: "asr" for an available auto-generated track. Transcript parameters. |
Transcript segments include text and start/end timestamps. Optional result fields depend on the video or channel, so ask your agent to flag missing data when making comparisons.
Connect your MCP client
Use https://mcp.hasdata.com/api/mcp?apis=youtube. The apis=youtube filter exposes the four YouTube tools.
For Cursor, add this to .cursor/mcp.json and replace the placeholder with your HasData API key:
{
"mcpServers": {
"youtube": {
"url": "https://mcp.hasdata.com/api/mcp?apis=youtube",
"headers": {
"x-api-key": "YOUR_HASDATA_API_KEY"
}
}
}
}
Clients with OAuth support can add the endpoint and sign in with HasData. Client-specific settings and authentication guidance are in the MCP documentation.
Cost and free tier
Each successful YouTube tool call costs 10 credits. HasData's free plan renews 1,000 credits monthly with no credit card required: up to 100 YouTube calls if you use those credits only for YouTube. Search pagination and additional transcripts consume separate calls. Check the current pricing for paid plans and their billing terms.
评论