SpockFetch: the page reader of Spock (https://spock.chat), an AI assistant.

What it does
  When a user asks Spock about something on the web, Spock reads the pages that question
  needs, on that user's behalf, in real time. It fetches one page per need. It does not
  crawl, build an index, or use pages to train models.

How to recognise it
  User-Agent: SpockFetch/1.0 (+https://spock.chat)
  Its requests are signed with Web Bot Auth (RFC 9421); the key directory is at
  https://agent.spock.chat/.well-known/http-message-signatures-directory
  Pages that need JavaScript are opened in a browser session in the requesting user's
  own sandbox. Those browser requests are not signed yet; they will be once the key
  is verified by the major bot-management providers.

What it respects
  robots.txt rules addressed to SpockFetch.
  Content-Signal: ai-input=no (the site's pages are not read as AI input).
  A challenge from the site that a real browser cannot clear: Spock stops and tells the
  user. It does not send that page through other networks.
  When Spock's own reader fails for a technical reason (not a refusal), a commercial
  fetch service may retrieve the page. That fallback is being retired.
  HTTP 402: Spock does not pay per page and does not read the page.

To block or limit SpockFetch
  Name it in robots.txt, e.g.
    User-agent: SpockFetch
    Disallow: /
  or send Content-Signal: ai-input=no.

Contact: louis@spock.chat