ytt-mcp
Enables users to fetch YouTube video transcripts via a single MCP tool, with support for multiple languages and OAuth-protected access.
README
ytt-mcp
A deliberately tiny, single-user MCP server with one tool: get_transcript. It runs on a Cloudflare Worker and protects the MCP endpoint with a standards-compliant OAuth flow backed by one password.
It follows the transcript retrieval flow used by jdepoix/youtube-transcript-api: load the watch page, call YouTube's InnerTube player API, prefer a manual caption track over an autogenerated one, then fetch and parse the timed-text XML. That project is the reference and is MIT licensed; this Worker is a small TypeScript implementation using only fetch.
Run locally
Install dependencies and create your local secret file:
pn install
cp .dev.vars.example .dev.vars
Set a long password in .dev.vars, then start the Worker:
pn dev
The MCP endpoint is:
http://localhost:8787/mcp
An unauthenticated request returns 401 and OAuth discovery information. The easiest way to test the complete browser login is with the MCP Inspector:
pn dlx @modelcontextprotocol/inspector
In the Inspector, select Streamable HTTP, enter http://localhost:8787/mcp, and connect. Its OAuth flow opens the ytt-mcp authorization page. Enter the password from .dev.vars.
Deploy
Create the production KV namespace:
pn exec wrangler kv namespace create OAUTH_KV
Replace the placeholder id under kv_namespaces in wrangler.jsonc with the returned ID (this repo is already configured with its production namespace). Set the production password as an encrypted Worker secret:
pn exec wrangler secret put MCP_PASSWORD
Deploy manually:
pn deploy
Pushes to master also deploy through .github/workflows/deploy.yml. Add these secrets under GitHub repository → Settings → Secrets and variables → Actions:
CLOUDFLARE_API_TOKEN— a Cloudflare API token with Workers Scripts: Edit and Workers KV Storage: Edit permissions.CLOUDFLARE_ACCOUNT_ID— your Cloudflare account ID.
The MCP_PASSWORD remains a Cloudflare Worker secret configured with Wrangler; do not add it to the GitHub workflow.
Configure your MCP client with:
https://ytt-mcp.<your-subdomain>.workers.dev/mcp
Do not put the production password in wrangler.jsonc, .dev.vars.example, or Git. Consider adding a Cloudflare rate-limiting rule for /authorize to reduce password-guessing attempts.
Tool
get_transcript
{
"video_id": "dQw4w9WgXcQ",
"languages": ["en"]
}
languages is optional and defaults to ["en"]. Codes are checked in priority order. For each code, a manually created transcript is preferred over an autogenerated transcript, matching the reference project.
The result is JSON text containing video/language metadata and snippets with text, start, and duration.
Authentication
@cloudflare/workers-oauth-provider handles OAuth discovery, dynamic client registration, PKCE, authorization codes, access tokens, refresh tokens, and bearer-token validation. OAuth data is stored in the OAUTH_KV namespace.
The custom /authorize page asks for one shared password and maps every successful login to the fixed user owner. There are no accounts or tenants. Existing access tokens remain valid after changing the password until they expire or are removed from KV.
YouTube can rate-limit or block cloud data-center IPs, and some videos require a PO token; those cases are returned as MCP tool errors.
ChatGPT compatibility patch
A two-line pnpm patch (in patches/) widens the OAuth library's CIMD handling so connectors that declare token_endpoint_auth_method: "private_key_jwt" (e.g. ChatGPT) are accepted and treated as public clients:
CIMD_ALLOWED_AUTH_METHODSgains"private_key_jwt"so the metadata document passes validation.- The negotiated auth method maps
private_key_jwt→"none", so the token exchange works.
Security implications: the client's JWT assertion is not verified — the flow is protected instead by PKCE (S256, enforced for public clients), single-use authorization codes stored in KV, and redirect_uri validation against the client's metadata document. This is a deliberate trade-off for a single-password personal server. Re-apply with pnpm install after any upgrade of @cloudflare/workers-oauth-provider.
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
Neon Database
MCP server for interacting with Neon Management API and databases
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.
E2B
Using MCP to run code via e2b.