Deploy and Host YouTube Transcript MCP with Railway

Give Claude, ChatGPT, or another MCP client a way to search YouTube and read its captions. YouTube Transcript MCP turns short caption fragments into readable, timestamped paragraphs with chapter headings and deep links back to the video. Ask about a whole talk, one section, or a specific topic, and get an answer you can verify.

About Hosting YouTube Transcript MCP

This template deploys the public GitHub repository as an HTTP MCP service. It provides a public HTTPS domain and a persistent volume for sign-in state. The server starts in a safe setup mode while credentials are missing; visit /healthz to see what remains. Railway's cloud IP is often blocked when fetching YouTube captions, so add a rotating residential proxy. GitHub sign-in keeps strangers from using your public endpoint and proxy bandwidth. After Railway gives you a domain, create a GitHub OAuth app, add your credentials, then connect https://YOUR-DOMAIN/mcp to your AI app. Railway hosting and the proxy may cost money. Running locally via stdio needs no proxy or GitHub OAuth app.

Common Use Cases

Summarize a lecture or interview with links to the moments behind each point.
Search a long video for a topic without sending the entire transcript to your model.
Browse a video's chapters before choosing what to read.
Find relevant videos on YouTube, then inspect their captions in the same conversation.

Dependencies for YouTube Transcript MCP Hosting

The remote service needs GitHub sign-in and a rotating residential proxy before it can answer transcript requests. The deploy itself starts in setup mode so you can configure both after Railway assigns a domain.

Deployment Dependencies

A GitHub account and OAuth app for access control. Set its callback to https://YOUR-DOMAIN/auth/callback.
A Webshare Rotating Residential plan, or another rotating residential proxy, because YouTube commonly blocks cloud IPs from caption endpoints. Webshare's free Proxy Server product is different; see the setup guide for the exact steps and a local option.
A Claude custom connector, supported ChatGPT plan, or another remote MCP client.

Read the five-step setup guide for the five Railway variables and a copy-paste prompt that lets your coding agent walk you through setup. Keep credentials in Railway Variables, not in an AI chat.

Why Deploy YouTube Transcript MCP on Railway?

Railway runs the HTTP service, provides its public HTTPS URL, monitors /healthz, and keeps OAuth registrations on a volume between deploys. This gives your AI app a stable, protected MCP endpoint without running a server on your laptop.