generated: '2026-09-07'
method: searched
source: https://www.drinkaccelerator.com/agents.md
docs: https://www.drinkaccelerator.com/llms.txt
limit_count: 0
rate_limits: []
note: >-
  An honest zero. Rate limiting is DOCUMENTED but never QUANTIFIED: /agents.md and
  /llms.txt both state "Respect rate limits. The MCP endpoint is rate-limited per IP.
  Back off on 429 responses." No numeric limit, no window, no burst, and no
  RateLimit-* or Retry-After header is published anywhere, so there is no limit to
  record as a row. Live probes of the MCP endpoint and the GraphQL endpoint returned
  no rate-limit response headers.
documented_but_unquantified:
- surface: mcp
  endpoint: https://www.drinkaccelerator.com/api/ucp/mcp
  scope: per-ip
  window: null
  limit: null
  burst: null
  status_on_exhaustion: 429
  response_headers: []
  guidance: Back off on 429 responses.
  source: https://www.drinkaccelerator.com/agents.md
cost_signal:
- surface: graphql
  endpoint: https://www.drinkaccelerator.com/api/2026-07/graphql.json
  field: extensions.cost.requestedQueryCost
  observed_value: 3
  observed_on: anonymous introspection probe
  note: >-
    The Storefront GraphQL API returns a per-request cost object in extensions. This
    is the only runtime consumption signal either surface emits, and it is a platform
    behaviour, not a published limit by this company - no throttleStatus was returned
    on the anonymous call.
x-evidence:
- url: https://www.drinkaccelerator.com/agents.md
  http_status: 200
- url: https://www.drinkaccelerator.com/api/ucp/mcp
  http_status: 200
  note: no RateLimit-* or Retry-After headers on the response
- url: https://www.drinkaccelerator.com/api/2026-07/graphql.json
  http_status: 200
  note: extensions.cost.requestedQueryCost = 3