generated: '2026-08-11'
method: probed
source: 'https://www.curehydration.com/llms.txt + live response headers on /products.json and /api/2026-04/graphql.json'
summary: >-
  Rate limiting is acknowledged but never quantified. The store's llms.txt tells agents the
  MCP endpoint is rate-limited per IP and to back off on 429, and gives no number, window,
  or burst. What IS available is a live cost signal on every response, which is the more
  useful runtime input for an agent — recorded below.
limit_count: 0
limit_count_note: >-
  Zero published numeric limits. This is an honest zero, not an unchecked field: llms.txt
  states the existence of a per-IP limit on /api/ucp/mcp without stating its value, and no
  Retry-After or RateLimit-* header was returned on any probed response.
documented: true
docs: https://www.curehydration.com/llms.txt

limits: []

runtime_signals:
  - header: shopify-complexity-score
    surface: 'GET /products.json'
    observed_value: '1490'
    probed: '2026-08-11'
    note: Per-response cost, not a remaining-quota counter.
  - header: shopify-complexity-score-v2
    surface: 'GET /products.json'
    observed_value: '149'
    probed: '2026-08-11'
  - field: extensions.cost.requestedQueryCost
    surface: 'POST /api/2026-04/graphql.json'
    observed_value: 3
    probed: '2026-08-11'
    note: >-
      The GraphQL response body returns the cost of the query just executed. Shopify's
      Storefront API normally also returns throttleStatus (maximumAvailable / currentlyAvailable
      / restoreRate); only requestedQueryCost was present on the anonymous probe.

headers_absent:
  - RateLimit-Limit
  - RateLimit-Remaining
  - RateLimit-Reset
  - X-RateLimit-Limit
  - X-RateLimit-Remaining
  - Retry-After
headers_absent_note: 'None of the standard rate-limit headers appeared on any 200 response probed on 2026-08-11.'

exhaustion:
  status_code: 429
  scope: per source IP
  surface: 'POST /api/ucp/mcp'
  behavior: 'Back off on 429 responses.'
  source: https://www.curehydration.com/llms.txt
  observed: false
  observed_note: Not reproduced — the pipeline does not deliberately exhaust a provider's limits.

gaps:
  - No numeric limit, window, or burst is published for any surface.
  - No RateLimit-* or Retry-After header on the responses probed, so an agent has no forward-looking budget signal — only a cost-of-last-request signal.