# YouTubeSearch

> Clean YouTube data for AI agents — search, transcripts, and summaries over one API, CLI, and MCP server, returned as agent-ready text.

YouTubeSearch exposes four operations: search YouTube, read video metadata and chapters, pull timestamped transcripts (captions when they exist, ASR when they don't), and fetch structured summary sections. Keyless requests serve search plus already-cached content; an API key (`Authorization: Bearer ys_live_...`) unlocks everything — the free tier includes 1,000 credits/month, no card. Base URL: https://api-production-4a11e.up.railway.app

## Docs

- [Documentation](https://youtubesearch-ai.vercel.app/docs): quickstart, authentication, all endpoints, error codes, MCP and CLI setup
- [Full API reference](https://youtubesearch-ai.vercel.app/llms-full.txt): the complete reference as one markdown file — enough to make a correct first call
- [OpenAPI spec](https://api-production-4a11e.up.railway.app/v1/openapi.json): machine-readable schema for every endpoint
- [MCP endpoint](https://mcp-production-f86c.up.railway.app/mcp): remote Streamable HTTP MCP server exposing the four operations as read-only tools

## Optional

- [Pricing](https://youtubesearch-ai.vercel.app/#pricing): keyless, free, and pro tiers with credit costs
- [Privacy](https://youtubesearch-ai.vercel.app/privacy): what we collect and what we don't
- [Terms](https://youtubesearch-ai.vercel.app/terms): beta terms of service