What it does
Given a YouTube video link or bare video ID, it returns the video's title, channel and full spoken transcript with timestamps as JSON. It accepts watch, youtu.be, shorts, embed and live links. Metadata comes from YouTube's official oEmbed endpoint. Transcripts are tried first with youtube-transcript-api and, if that fails, through Invidious instances.
Use it when an AI agent needs the content of a video it cannot watch. It is exposed as a plain HTTP endpoint (POST /read, with a health check) and as a single MCP tool, youtube_reader(url), for clients such as Claude Desktop, Cursor and Windsurf.
To run it, push the project folder to a GitHub repository and deploy it on Railway's free tier; the platform reads requirements.txt and Procfile automatically. Metadata is free. MCP transcript calls are priced per call, and the plain HTTP API is free while in beta. Per-call pricing is set in mcpize.yaml and the MCPize dashboard, and the publisher's PAYMENTS.md covers wallet and payment setup.
Limitations and requirements
YouTube rate-limits automated requests from datacenter IPs, so transcript fetches can fail intermittently from cloud servers; metadata is not affected. The publisher suggests a residential proxy as a third fallback, at about $3 per 1,000 transcripts. MCP transcript calls cost $0.02 each, paid in USDC on Base via x402. A $19/month tier is planned but not yet available.
Safety
Our team has not reviewed this resource. Only automated rules checked it; resources with high-risk signals are not published. These are signals, not a guarantee that installing it is safe. Try it in a sandbox project first.
- Runs code or changes an assistant's behaviour
Sources
- Official MCP Registry first seen 13 Nov 2026 · last seen 13 Nov 2026 · its page on the source
YouTube Reader
Turn any YouTube video into text: transcript, title, channel, timestamps.
v1.0.0As shown on registry.modelcontextprotocol.io when we last read it