meta-externalagent identifies with a user agent beginning meta-externalagent and follows robots.txt, so you control it the standard way. Its job is collecting public data for Meta's AI models and indexing content.
It is worth separating from Meta-ExternalFetcher, the agent that retrieves a specific link when a user shares or acts on it. User-initiated fetches like that may not honor robots.txt, so a robots rule governs the training crawler more reliably than the fetcher.