generated: '2026-09-19'
method: probed
source: live unauthenticated responses from https://allagents.app (GET /agents, /search, /agent/<slug>, POST /a2a) observed 2026-09-19; https://allagents.app/api and /llms.txt read for documented limits
limit_count: 0
summary: >-
  No rate limit is documented anywhere the provider publishes (/api, /llms.txt, homepage) and none was
  signalled on any live response — headers observed were Server (nginx/1.28.3), Date, Content-Type,
  Transfer-Encoding and Connection only. An honest zero. The two documented caps are RESPONSE-SIZE caps,
  not rate limits, and are recorded below as such so nobody mistakes them for a quota.
headers:
  observed: []
  documented: []
status_on_exhaustion: null
scopes: []
size_caps:
  - surface: GET /search
    param: limit
    cap: 25
    note: '"scored matches, small responses (limit ≤ 25)" — limit=100 returned 25 on 2026-09-19.'
  - surface: GET /agents, GET /category/<specialty>
    page_size: 10
    note: '"the book, 10 per page, newest first" — page size is fixed, not configurable.'
notes: >-
  Absence of a 429 or Retry-After on a handful of probes is not proof that no throttle exists at the edge;
  it is proof that none is published or signalled. Agents should pace themselves conservatively and treat
  any future 429 as authoritative.