What it does
meta-externalagent is Meta's training crawler. Content is copied into datasets used to train foundation models. Blocking costs you nothing in traffic today; allowing is a donation of your content to model weights with no citation or referral in return.
Llama training + product improvement crawler. Observed live in our own logs (a production site we operate: meta-externalagent/1.1, 272 hits/24h on 2026-05-10) and honored our robots/403 combination. Lowercase token in the wild.
Full UA: not captured in our July 2026 roster review — match log hits on the token meta-externalagent (case-insensitive substring) and verify by IP before acting.
How to verify a hit is really meta-externalagent
Verification method (July 2026): IP ranges discoverable via Meta ASN 32934 (RDAP/whois); docs page fetched live 2026-08-22
No per-crawler IP list is published; the practical check is network ownership — look up the hit's source IP via RDAP/whois and confirm it announces from the vendor's network (ASN) rather than a residential proxy or cloud range.
User-agent strings are freely spoofed, so a UA match alone proves nothing. Full step-by-step method: verify by IP.
Allow or block in robots.txt
Match the exact token meta-externalagent in robots.txt:
# Block meta-externalagent site-wide
User-agent: meta-externalagent
Disallow: /# Explicitly allow meta-externalagent
User-agent: meta-externalagent
Allow: /For the allow-search-block-training combined pattern (and the Allow-directive gotcha that silently kills carve-outs), see the robots.txt guide. robots.txt governs whether meta-externalagent may fetch; llms.txt is the separate, curated map AI assistants read once allowed in.
Vendor documentation: https://developers.facebook.com/docs/sharing/webmasters/web-crawlers
The Meta family
Meta operates 4 distinct tokens in this roster, and they do not block each other: a robots.txt line for meta-externalagent does nothing to its siblings — each needs its own User-agent: entry (the robots.txt guide has combined patterns).
| Token | Role | robots.txt |
|---|---|---|
| meta-externalfetcher | User action | ignores robots.txt |
| meta-webindexer | AI search | claims robots compliance |
| meta-externalads | Training | claims robots compliance |
Frequently asked questions
What is meta-externalagent?
meta-externalagent is Meta's training crawler. Content is copied into datasets used to train foundation models. Blocking costs you nothing in traffic today; allowing is a donation of your content to model weights with no citation or referral in return.
Does meta-externalagent respect robots.txt?
Meta claims meta-externalagent respects robots.txt, but compliance is not independently confirmed.
How do I block meta-externalagent?
Add 'User-agent: meta-externalagent' followed by 'Disallow: /' to your robots.txt. Use the exact token — substring variants of other tokens will not match.
How do I verify meta-externalagent traffic by IP?
No per-crawler IP list is published; look up the source IP via RDAP/whois and confirm it announces from Meta's network before trusting the user agent.
Related bots
Same vendor, then other training crawlers — the full directory profiles all 34.