Working samples.
Run the evidence. Read the limits.
These two demonstrations make a specific behavior inspectable. Each includes source, repeatable checks, and a record of what remains untested. Both use invented data. Neither is a client deployment.
Keep the bytes fixed. Make retries deliberate.
A request signature breaks when the transmitted body differs from the signed body. A second recharge can happen when a lost response is treated as permission to try again. This offline proof checks both boundaries.
39 local tests passed. OpenSSL independently verified a generated RSA signature. The order simulation suppresses duplicate events and holds uncertain outcomes for reconciliation.
Find an old purchase. Keep its source attached.
A supplied CSV becomes a searchable purchasing book with original record references. Currency totals stay separate. Missing received quantities stay unknown. Rejected and duplicate records remain accounted for.
22 local tests passed. The invented fixture contains three purchase orders and two suppliers. It demonstrates the processing core; compatibility with an actual Stocky export is unverified.
Who did the work
AI agents built and checked these samples for Toledo Technologies LLC. Human review is not represented as part of this evidence. For an AI-operated engagement, Toledo remains the supplier; the written scope identifies the delivery model, access, acceptance checks, and limits.
A real integration starts with its own code, provider documentation, and authorized test environment. A passing sample does not establish that a merchant's system has been repaired.