generated: '2026-09-13'
method: searched
source: https://hearthnhome.com/llms.txt
note: >-
  The only published rate-limit statement HNI serves is one prose sentence in the agent instructions at
  hearthnhome.com. No numeric limit, no window, no burst allowance and no response headers are
  documented, so limit_count is an honest zero.
limit_count: 0
limits: []
documented_behaviour:
  scope: per-IP
  statement: 'The MCP endpoint is rate-limited per IP. Back off on 429 responses.'
  exhaustion_status: 429
  source: https://hearthnhome.com/llms.txt
headers:
  documented: []
  note: >-
    No X-RateLimit-*, RateLimit-* or Retry-After header is documented anywhere HNI publishes.
  observed:
    - header: shopify-complexity-score
      example: '370'
      note: >-
        Per-call cost score returned on every response. Undocumented, but it is the only quantitative
        budget signal the endpoint emits — an agent can track cumulative complexity even though it is
        never told the ceiling.
    - header: shopify-complexity-score-v2
      example: '37'
    - header: x-request-id
      example: 98c301df-79d4-4a8d-b6e0-546c589fb0c7-1789302129
      note: Correlation identifier, useful when reporting a throttle.
  observed_probe:
    url: https://hearthnhome.com/api/ucp/mcp
    method: POST tools/list
    http_status: 200
    probed: '2026-09-13'
observed: []
observed_note: >-
  A single unauthenticated tools/list call returned HTTP 200 on 2026-09-13 carrying complexity-score
  headers but no RateLimit-*/Retry-After header. The endpoint was not load-probed and no 429 was
  deliberately triggered.
gap: >-
  Rate limiting is asserted but unquantified and unsignalled. This is the weakest runtime contract on
  the surface: an agent cannot plan concurrency, and cannot distinguish a throttle from an outage
  without a Retry-After.