Do I need an API key or account?

No. There's no signup, no dashboard, and no key to provision. Authentication is the payment itself — a signed USDC transfer via the x402 protocol.

How do agents discover the price before paying?

Call any endpoint without a payment header and it returns HTTP 402 with the price quote. Or read https://api.sumvid.app/.well-known/x402.json directly — the x402 discovery manifest lists current prices, network, and payTo for every endpoint up front.

What happens if a call fails?

You are not charged. Payment is verified before the work runs but only settles on a successful response — a bad URL or a video with no captions returns an HTTP error and nothing settles.

Can I use this from an agent framework, not just raw HTTP?

Yes — the same three capabilities are exposed as MCP tools over streamable-http, so any MCP-capable agent framework can call them as native tools instead of hand-rolling HTTP requests.

For agents & developers

YouTube summary API for AI agents

Sumvid's YouTube summary API is an HTTP API and an MCP server that summarize YouTube videos, fetch transcripts, and summarize whole channels — built for AI agents to call directly. Every call is paid per use in USDC via the x402 protocol: no API keys, no account, and no subscription to set up first.

Endpoints

Three capabilities

No prices below — per-call prices are set server-side and can change, so the x402 discovery manifest is the machine-readable source of truth for current pricing, network, and payTo.

POST /v1/summarize: Returns an AI-generated summary of a single YouTube video. What an agent wants when it needs the gist of a video — key takeaways, not the full transcript — to reason over or hand back to a user.
POST /v1/transcript: Returns the full plain-text transcript of a single YouTube video. What an agent wants when it needs to quote, search, or run its own analysis over the source text instead of a pre-summarized version.
POST /v1/channel/new: Returns summaries for a channel's recent videos in one call. What an agent wants when it's monitoring a channel and needs to catch up on everything new at once, instead of paging through videos one at a time.

How it works

How x402 works

1. Request. Call an endpoint like any HTTP API — no key, no auth header.
2. 402 with payment requirements. Without payment, the response is HTTP 402 with the price, network, and payTo address the endpoint requires.
3. Retry with payment. Retry the same request with a signed USDC payment header. The payment is verified before the work runs, and only settles once the call succeeds — a failed call is never charged.

MCP

Native MCP tools

The same three capabilities are exposed as MCP tools over streamable-http, so agent frameworks that speak MCP can call them as native tools instead of hand-rolling HTTP requests — same x402 payment model, same no-signup access. Agent-readable index: https://api.sumvid.app/llms.txt.

Discovery

Find current prices before you call

The discovery manifest is plain JSON — no wallet or payment library needed just to check pricing:

curl -s https://api.sumvid.app/.well-known/x402.json | jq .
# → x402 discovery manifest: current per-call prices, network,
#   and payTo address for every endpoint below, machine-readable.

When not to use this

If you're a person who wants to skim a video rather than an agent calling this programmatically, use the Sumvid app instead. Sumvid's free tier is cheaper than paying per call for casual, one-off use — this API exists for agents and integrations that need programmatic access, not as the default way for a human to get a summary.

See also: Tools & API for a shorter, in-product index of these same endpoints.

FAQ

Common questions

Do I need an API key or account?: No. There's no signup, no dashboard, and no key to provision. Authentication is the payment itself — a signed USDC transfer via the x402 protocol.
How do agents discover the price before paying?: Call any endpoint without a payment header and it returns HTTP 402 with the price quote. Or read https://api.sumvid.app/.well-known/x402.json directly — the x402 discovery manifest lists current prices, network, and payTo for every endpoint up front.
What happens if a call fails?: You are not charged. Payment is verified before the work runs but only settles on a successful response — a bad URL or a video with no captions returns an HTTP error and nothing settles.
Can I use this from an agent framework, not just raw HTTP?: Yes — the same three capabilities are exposed as MCP tools over streamable-http, so any MCP-capable agent framework can call them as native tools instead of hand-rolling HTTP requests.