SERP ENGINE

YouTube

YouTube in three tools, live via SerpAPI. Search returns videos plus channels, playlists and movies; each video’s id feeds the video tool — details, related videos, comments and replies — and the transcript tool, which pulls complete transcripts in any available language, including auto-generated captions. That last one turns any talk, lecture or interview into text your agent can actually read.

3 tools ~2–8 s cache: 5 min SerpAPI youtube / youtube_video / youtube_video_transcript
Connect

Thirty seconds, any client.

claude mcp add --transport http youtube https://mcp.wondel.ai/youtube/mcp

One command in your terminal — the server is available in every Claude Code session.

This server works with Claude, Claude Code, Claude Desktop, Claude Cowork, Cursor, VS Code, Windsurf — and any other MCP client. Stdio-only clients connect through npx mcp-remote.

The Tools

3 tools, fully typed.

Parameters below are transcribed from the server’s own zod schemas — what you see is what the tool validates.

youtube_search Search YouTube: videos, channels, playlists, movies and related searches. The sp token carries filters and pagination from a prior result.
Param Type Description
search_query * string Search query — same syntax as the YouTube search box.
sp string Filter/pagination token from a prior result.
gl string Country code, e.g. "us", "de".
hl string Language code, e.g. "en", "pt-br".
no_cache boolean
= false 		Force a fresh SerpAPI fetch, bypassing this server’s result cache.
zero_trace boolean
= false 		Enterprise-only ZeroTrace mode — SerpAPI stores no search params or metadata.
youtube_video One video by id: details, related videos, comments and replies, paged with next_page_token.
Param Type Description
v * string YouTube video id — from a youtube_search video_results entry.
next_page_token string Pages through related videos, comments or replies.
gl string Country code, e.g. "us".
hl string Language code, e.g. "en".
no_cache boolean
= false 		Force a fresh SerpAPI fetch, bypassing this server’s result cache.
zero_trace boolean
= false 		Enterprise-only ZeroTrace mode — SerpAPI stores no search params or metadata.
youtube_transcript A video’s transcript. Pick the track by language, title or type ("asr" = auto-generated).
Param Type Description
v * string YouTube video id — from a youtube_search result.
language_code string
= en 		Transcript language code, e.g. "es".
title string Track title, when multiple tracks exist.
type string Track type, e.g. "asr" for auto-generated captions.
no_cache boolean
= false 		Force a fresh SerpAPI fetch, bypassing this server’s result cache.
zero_trace boolean
= false 		Enterprise-only ZeroTrace mode — SerpAPI stores no search params or metadata.
Try It

Prompts that just work.

Find the most-viewed talks about Durable Objects and pull the transcript of the best one.

youtube_search + youtube_transcript

Summarize the top comments under this video — what do viewers keep asking?

youtube_video

Get the Spanish transcript of this keynote and translate the Q&A section.

youtube_transcript

What comes back

{
  "source": "serpapi/youtube_video_transcript",
  "fetchedAt": "2026-07-02T14:02:11.312Z",
  "query": { "v": "dQw4w9WgXcQ", "language_code": "en" },
  "transcript": [
    { "start": 0.0, "duration": 4.2, "text": "We're no strangers to love…" },
    { "start": 4.2, "duration": 3.8, "text": "…" }
  ]
}

Abridged real response. Every payload carries the same envelope — source says where the data came from, fetchedAt says when, query echoes what you asked. Compact JSON in one text block, exactly how agents like it.

Add YouTube
to your agent.

mcp.wondel.ai/youtube/mcp

Authless · Read-only public data · Streamable HTTP at /mcp, legacy SSE at /youtube/sse