generated: '2026-08-23'
method: searched
source: https://kite-hill.com/llms.txt
docs: https://shopify.dev/docs/api/usage/limits
description: >-
  Kite Hill publishes no numeric rate limits. Its own agent instructions state that the MCP
  endpoint is rate-limited per IP and that agents must back off on 429; the GraphQL surface is
  cost-limited and reports the cost of every query in the response body rather than in
  headers.
limit_count: 0
limits:
- surface: mcp
  endpoints: [https://kite-hill.com/api/ucp/mcp, https://kite-hill.com/api/mcp]
  scope: per-ip
  window: not published
  limit: not published
  burst: not published
  exhaustion_status: 429
  quote: 'Respect rate limits. The MCP endpoint is rate-limited per IP. Back off on 429 responses.'
  source: https://kite-hill.com/llms.txt
- surface: graphql
  endpoint: https://kite-hill.com/api/2026-04/graphql.json
  scope: query-cost
  window: not published
  limit: not published
  signal: extensions.cost.requestedQueryCost on every response
  observed:
  - query: '{shop{name}}'
    requestedQueryCost: 1
  - query: '{__schema{queryType{name}}}'
    requestedQueryCost: 3
  - query: 'collections + products + pages + shop composite'
    requestedQueryCost: 38
  exhaustion_status: 200 with a THROTTLED error in errors[] (documented by the platform; not observed)
  docs: https://shopify.dev/docs/api/usage/limits
response_headers:
  observed: []
  note: >-
    No X-RateLimit-*, RateLimit-* or Retry-After headers were returned on any probed response
    — the runtime signal an agent actually needs is absent on both surfaces. On GraphQL the
    cost is in the body; on MCP there is no signal until the 429 arrives.
x-evidence:
- fetched: '2026-08-23'
  url: https://kite-hill.com/llms.txt
  http_status: 200
- fetched: '2026-08-23'
  url: https://kite-hill.com/api/2026-04/graphql.json
  http_status: 200
  note: extensions.cost present on every response; no rate-limit headers