MEGURO / FIDELITY
Measured compatibility

Same grammar. Named boundaries.

Meguro practice stores speak the Shopify Admin shapes we have harvested from real Shopify, then say plainly where the current surface stops.

43/55 Admin GraphQL checks pass 22/22 Storefront checks pass 5,945 API-call demo year Measured June 2026
Coverage

The declared surface.

43/55Admin GraphQL checks pass

The declared v1 subset covers Products, variants, Orders, inventory items, Customers, and read-only fulfillment-order context used by the launch scenarios.

22/22Storefront checks pass

Storefront reads are covered. The shopper endpoint models cart and checkout to the buyer handoff. Payment placement and broad Storefront mutation families are outside the launch claim.

Method

How we know.

Harvest

Meguro compares responses, status codes, and error bodies against real Shopify. The WIRE-A, WIRE-B, R-5, and R-7 lineage is the proof path behind the public number.

Inheritance

The grammar is proven once per driver against a real store. Practice-store users inherit that proof instead of spending hours on a dev-store setup loop for every run.

Failure mode

Unsupported calls fail loudly with named errors or userErrors. Meguro does not silently pretend a missing Shopify shape worked.

Speed

Measured, not promised.

DEMO-QF measured the launch reference workload — the reference-workload timing, measured July 7, 2026 — as 5,945 Shopify API calls, 372 reads, 5,572 writes, 1 blocked call, 365 store days, 24 m 31.5 s sequential. One call at a time, no rate limits.

WorkloadMeguro measurementReal dev-store comparison
Light compatibility agentAbout 1,500 calls/year; typically 10-20 minutes depending on the agent.Throttle and round-trip time are the agent's wall-clock floor.
Returns triage reference5,945 calls in 24 m 31.5 s, sequential — the reference-workload timing, measured July 7, 2026.Its own calls would be roughly an hour after setup, depending on round trips and backoff.
Manufacturing the yearThe store year is computed by Meguro and then exercised by the agent.~50-70,000 setup API calls, a hard 2-hour throttle floor, realistically a day, and real time still does not pass.
Exam lane

Shopify's answers, recorded.

The Exam replays the recorded requests on your Shopify development store and records Shopify's answers.

It proves only what Shopify returned for those requests on that development store. It does not decide whether the agent was right; the shared call index aligns the Practice record with the redacted public transcript.

Boundaries

What this does not claim.

Meguro reports pre-tax dollars. It is not a payments, tax, invoice, or settlement engine.

The practice world models demand and operations: orders, returns, stock movement, fulfillment pressure, and agent decisions. It does not claim future merchant demand.

Throttle modes are compatibility tools. The default standard-1000 practice lane uses Meguro’s modeled request-cost bucket; it is not exact Shopify rate-limit parity.