Clean YouTube data for agents.
Search, video metadata, timestamped transcripts, and structured summaries. Four operations return plain fields and markdown, so an agent reads only the range it needs.
- REST API
- MCP server
- Command line
Four operations, one clean contract
Instead of loading an entire video into context, an agent asks for exactly the field, section, or time range it needs. Every response is plain structured data — no player, no scraping, no HTML to unpick.
- Interfaces
- REST API · MCP server · CLI
- Operations
- search · metadata · transcript · summary
- Formats
- JSON fields · Markdown
- Access
- API key required — free tier, no card
The endpoint index
One call per operation. Cost is metered in credits — a summary reads more than a search, and nothing else surprises you.
From request to clean data
A video goes in as an id or a query. What comes back is aligned, structured, and ready to read — the messy middle is handled.
- fields
- text
- markdown
A single exchange
One request for a transcript range. The response carries only the segments asked for.
GET /v1/transcript?id=dQw4w9WgXcQ&range=01:30-03:00{
"id": "dQw4w9WgXcQ",
"range": "01:30–03:00",
"language": "en",
"segments": [
{ "t": "01:32", "text": "The agent fetches only the range it needs," },
{ "t": "01:47", "text": "so a 90-minute talk costs a handful of tokens." }
]
}Two tiers
Take a free key for 1,000 credits a month — same speed as pro. Move to Pro when an agent runs in production.
Start with 1,000 credits a month.
Free key, 1,000 credits/mo, no card — same speed as pro