MODEL API GATEWAY

PackyAPI

A model API gateway with documented OpenAI-style access and model-specific setup guides. Authenticated requests and billing remain untested.

Invitation code: Fspr · Included in this registration link.
Referral link: this registration URL contains the site owner's affiliate/referral identifier. Commission terms and reader benefits have not been verified.

Public documentation reviewed · 2026-10-09 · Documentation review · Hands-on test pending

Prepared by EcomAgentHub editorial · Method and evidence limits

Official PackyAPI homepage screenshot; not an EcomAgentHub test · Image source ↗

Fixed cases and current execution status

Prepared API-gateway evaluation cases, all unexecuted. These cases evaluate a selected configuration, not every routed model or a connected store. No account, secret token, model output, measured cost or performance score is published as a result

Prerequisites: A human must confirm applicable terms, regional eligibility, an approved account and model/token group, available credits and a spending limit. Use synthetic inputs and keep keys in local environment variables. No paid request is authorized by this prepared plan.

Verify one supported model, protocol and token group · Not executed

A human-approved model name, token group, client and connection address, recorded from current official documentation. No token is included in the public record.

After account, terms and budget approval, submit one small synthetic request through a documented protocol. Retain the first request projection and full response or error with credentials removed. Record the configured model and any returned model identifier without treating them as proof of upstream identity.

Expected: The chosen configuration returns an inspectable response or a documented failure. Access eligibility, model selection and protocol support are reported separately.

Model, token group, client and protocol are recorded before execution.
The first response or failure is retained without exposing a secret.
Reported model fields are distinguished from independently verified provider identity.

Keep a synthetic product description within supplied facts · Not executed

Synthetic CedarDesk desk mat: one charcoal mat, 60 x 30 cm, 100% polyester felt. No certification, medical, sustainability, waterproof, customer-count or sales claim is approved.

Using the approved text configuration, draft a 60–90-word description and exactly three bullets using only these facts. Do not publish or claim a store action. Retain the first output and every attempted request, including errors.

Expected: A description within the requested length and exactly three bullets preserve quantity, dimensions, material and colour without adding unsupported claims.

Description length is 60–90 words and there are exactly three bullets.
Every supplied product fact is preserved; unsupported claims remain absent.
No publication, account change or connected-store action is asserted.

Record supported streaming and validation failures · Not executed

The same approved text configuration and a synthetic request. Streaming is used only when its documentation permits it; one non-secret invalid-model input may be prepared to inspect error handling.

Within the agreed request limit, retain complete stream events, termination information and a validation error where permitted. Record interruptions and unsupported streaming as limits; do not evade quotas, probe unrelated endpoints or retry until a result looks successful.

Expected: The record distinguishes a complete response from partial output, interruption and rejected configuration, including any failed requests without output files.

Streaming support is confirmed for this configuration before the request.
First events, completion state and errors are retained with secrets removed.
No-output attempts remain in the attempt ledger and are not counted as successful completions.

Reconcile request usage with actual gateway charges · Not executed

Human-approved spending cap; actual pre-test and post-test account usage records for the planned requests. No payment details or account identifiers are included in public evidence.

Record actual token/usage fields, credits debited, billed currency, selected model/group and charges for each success or failure. Explain how paid credits are allocated to this test. Keep unmeasured quantities unknown and stop at the agreed cap.

Expected: Measured charges can be reconciled to the retained attempts, or the record explicitly states which cost components cannot be determined.

Costs come from actual gateway records, not assumed upstream prices or an advertised discount.
Attributable failed-request costs are included and missing usage remains unknown.
Cost per usable output is calculated only with a measured cost basis and usable count; zero usable outputs shows No usable output.

Download this exact case plan · Who prepares and reviews these records

Who is it for?

Developers evaluating model access for their own ecommerce applications, such as fact-constrained product copy or internal workflows. A custom application and its permissions are separate from the gateway.

What the product advertises

The official homepage advertises unified model access, usage monitoring and routing controls
Official guides describe OpenAI-style requests and selected CLI/client configurations
Model-specific text and image guides specify token groups and endpoint requirements

Platform fit

Model APIs for custom applications · Native store connections unverified. Platform tags can describe a native connection, marketplace data, or exported content. The access and limitation notes below explain what to confirm.

Access and setup

PackyAPI account, available credits, an API token and a client configuration supported by the selected model and token group. The public registration page offers Google and username registration. No registration, authenticated API request or paid test was completed for this listing. Confirm the current availability and supported workflows in the official product documentation before choosing a plan.

Pricing notes

The quick-start documentation describes buying credits before creating an API token. The commercial terms refer to current model rates and possible prepayment. Model, token-group and request charges need a current account-level check; no free entitlement, numeric rate, discount, failed-request charge or actual cost per usable output has been verified here.

Official pricing & allowances ↗

Limits to check

This is model-access infrastructure; native Amazon, TikTok Shop or Shopify connections and store actions have not been verified
Protocol and client support depend on the model and token group. The group guide restricts third-party access for some groups; do not assume every listed model works with every client
The group guide describes some channels as reverse-engineered. Upstream authorization and the actual identity of responding models have not been independently verified
The GPTImage guide specifies the Images API, one image per request and no transparent-background or streaming support for that documented configuration; these are vendor-documented limits, not measured behavior
The homepage and quick-start guide display different connection addresses. Confirm the current address for the chosen client and token group rather than copying a generic URL
Output quality, streaming completion, errors, actual billing, latency and reliability remain untested. Advertised availability and cost advantages are not our test conclusions
Commercial and upstream terms, supported regions and data handling require review for the intended use. No independent privacy, security or commercial-rights assurance is established by this listing

How to evaluate it

Use one real task and a written set of success criteria. Check output accuracy, source traceability, setup time, and total cost. Review any proposed store action before enabling it. This listing summarizes public information; it does not establish performance in your store.

Sources and commercial relationship

Official product source ↗ · Reviewed 2026-10-09.

Referral link: this registration URL contains the site owner's affiliate/referral identifier. Commission terms and reader benefits have not been verified. There is no paid placement on this listing. Read our editorial approach.

Official PackyAPI homepage · Reviewed 2026-10-09. Rendered public product positioning and advertised model-access controls; no uptime or performance claim independently tested.
Official quick-start guide · Reviewed 2026-10-09. Account, credit, API-token and client-configuration prerequisites.
Model and token-group guide · Reviewed 2026-10-09. Group-dependent model access, client restrictions and vendor-described channel sourcing.
Kimi Code protocol guide · Reviewed 2026-10-09. Selected-model examples for Chat Completions, Responses and Anthropic Messages; not a universal compatibility claim.
GPTImage configuration guide · Reviewed 2026-10-09. Images API and documented image-parameter limits for the specified configuration.
Commercial terms · Reviewed 2026-10-09. Rendered terms covering fees, third-party policies, use restrictions and a linked data-processing addendum; no legal or privacy guarantee inferred.

Evidence update history

2026-10-09: documented feature, access and source review.