Start here
For bots and AI products
Ask which sites allow retrieval, what they consent to, and which URLs are worth crawling. Anonymous reads are free at 30 per minute — you only need a key to go faster.
Mint a key
Sign in and create a bc_live_ secret. It is shown once, so store it when you see it.
Create an API key →Call it
Send the secret as a Bearer token to /v1/search, /v1/site/{domain} or the MCP endpoint.
Endpoints and examples →No key yet? Every /v1 read works anonymously at 30 requests a minute.
For site owners
A listing states what bots may do with your pages. Nothing is listed until you prove you own the origin, and proof is re-checked on a schedule — a card cannot outlive the site behind it.
Probe the origin
Check robots.txt, llms.txt and the well-known path before you publish. The probe reports exactly what is missing.
Probe a domain →Prove you own it
Publish a DNS TXT record botcentral-verify=<token>, or a plain-text /.well-known/botcentral.txt. HTML at that path is rejected.
Listing rules →Publish the card
Declare consent — retrieve, train, act, cite — and the URLs you want crawled. Publisher products can list client sites over the API.
Publish a card →Retrieve is not train. Consent is split, and training stays off unless you turn it on.
The card format, the scoring, and the rate limits are all published. Nothing here changes without a contract version.