Trending repos · x402

GitHub trending API for AI agents

gitBeacon is a GitHub trending API built for AI agents and developers: it serves the raw enriched trending-repo rows from a daily scan — stars, forks, language, topics, license, author, and README excerpt — alongside a daily LLM-analyzed digest of what those repos add up to. Every paid call settles in USDC on Base over the x402 protocol, with no accounts and no API keys, so an agent can pull structured trending data directly and reason over it.

Endpoints

The free /v1/digests/latest and frozen /v1/sample reads show the digest shape before you pay.

How agents pay

Call a paid endpoint with no payment and gitBeacon answers HTTP 402 with the exact price. Your agent signs an EIP-3009 USDC authorization on Base and retries with the payment header; the trending rows return after settlement. You are never charged for an error.

$ curl "https://api.gitbeacon.dev/v1/repos?language=rust"
HTTP/2 402 Payment Required
# agent signs a USDC payment on Base and retries
HTTP/2 200 OK

FAQ

Is there a GitHub trending API with no API key?
Yes. gitBeacon serves trending-repo data and digests over plain HTTP with no API key or account. The digest feeds are free reads; the /v1/repos trending-repo endpoint and /v1/digests history settle per call over x402.
What does the trending-repos endpoint return?
GET /v1/repos returns the raw enriched trending-repo rows from the latest daily scan — stars, forks, language, topics, license, author, and a README excerpt — and is filterable by language.
How does an agent pay for trending data?
A paid call answers with HTTP 402 and the price — top trending repos are $0.01 per call. The agent signs an EIP-3009 USDC authorization on Base and retries with the payment header. No keys, no subscription.
How often does the trending data refresh?
gitBeacon runs a daily trending scan (24h cadence). Poll the free /v1/index feed to detect a new digest; paid rows carry scanDate/digestDate provenance so freshness is always explicit.
Built for agents

Discovery, machine-first