generated: '2026-08-17'
method: probed
source: >-
  https://lesechappees.arlettie.com/llms.txt (documented) plus response headers observed on
  /api/ucp/mcp, /api/2026-04/graphql.json and /products.json
note: >-
  Arlettie publishes a rate-limit RULE but no rate-limit NUMBER. llms.txt and robots.txt tell an agent
  the MCP endpoint is rate-limited per IP and to back off on 429, and every response carries a
  cost-accounting header — but no ceiling, window, remaining count or reset time is published anywhere.
  An agent can therefore see what a call cost but cannot compute how many it has left.

limits:
- scope: per-ip
  surface: https://lesechappees.arlettie.com/api/ucp/mcp
  window: null
  limit: null
  burst: null
  documented: true
  quote: >-
    "Respect rate limits. The MCP endpoint is rate-limited per IP. Back off on 429 responses."
  source: https://lesechappees.arlettie.com/llms.txt
  note: existence is documented; the value is not
- scope: per-query-cost
  surface: https://lesechappees.arlettie.com/api/2026-04/graphql.json
  window: null
  limit: null
  documented: false
  observed: >-
    Responses carry a cost object in the GraphQL extensions
    ("extensions":{"cost":{"requestedQueryCost":3}}) and a shopify-complexity-score header. Neither a
    bucket size nor a restore rate is published on this store.

response_headers:
- header: shopify-complexity-score
  meaning: cost accrued by this request
  observed_values:
    'POST /api/ucp/mcp tools/list': 37
    'POST /api/mcp tools/list': 50
    'POST /api/2026-04/graphql.json {shop{name}}': 0
    'GET /products.json': 573
  standard: false
- header: shopify-complexity-score-v2
  meaning: successor cost accounting header, returned alongside the original
  observed_values:
    'POST /api/ucp/mcp tools/list': 37
    'POST /api/mcp tools/list': 50
    'POST /api/2026-04/graphql.json {shop{name}}': 10
    'GET /products.json': 573
  standard: false
- header: x-request-id
  meaning: correlation id for support/debugging on a throttled call
  standard: false
- header: server-timing
  meaning: processing and db duration, edge and servedBy — latency attribution
  standard: false

absent_headers:
- RateLimit
- RateLimit-Policy
- X-RateLimit-Limit
- X-RateLimit-Remaining
- X-RateLimit-Reset
- Retry-After
absent_note: >-
  None of the IETF draft RateLimit fields nor the legacy X-RateLimit-* family nor Retry-After appeared
  on any 200 response from any surface. Retry-After may well be present on an actual 429; that was not
  verified, because provoking a 429 would mean deliberately abusing a live consumer storefront.

exhaustion:
  status_code: 429
  documented: true
  source: https://lesechappees.arlettie.com/llms.txt
  retry_after_header: unknown
  observed: false
  note: >-
    429 is the documented exhaustion code. No 429 was observed during this pass and none was induced.

graphql_cost_envelope:
  path: extensions.cost.requestedQueryCost
  observed: 3
  note: >-
    In-body cost accounting, which is the more useful of the two signals here because it is returned in
    the payload an agent already parses.

limit_count: 0
limit_count_note: >-
  Zero published numeric limits. The rate-limit posture is documented in prose and instrumented with a
  cost header, but there is no number an integrator can plan against.

x-evidence:
  fetched: '2026-08-17'
  probes:
  - url: https://lesechappees.arlettie.com/api/ucp/mcp
    http_status: 200
    headers_observed: [shopify-complexity-score, shopify-complexity-score-v2, x-request-id, x-shopify-ucp-mcp-api-version]
  - url: https://lesechappees.arlettie.com/api/2026-04/graphql.json
    http_status: 200
    headers_observed: [shopify-complexity-score, shopify-complexity-score-v2, x-shopify-api-version]
  - url: https://lesechappees.arlettie.com/products.json
    http_status: 200
    headers_observed: [shopify-complexity-score, shopify-complexity-score-v2]
  - url: https://lesechappees.arlettie.com/llms.txt
    http_status: 200