Research evidence and limits
Desk research conducted 28 September 2026. Sources were inspected as public documents. No competitor accounts were tested, merchants interviewed or keyword volumes measured.
This register separates published evidence from the product decisions we are proposing. Pages linked here carry the substantive summaries; dates and limitations matter when revisiting them.
| Source | What it supports | What it cannot establish |
|---|---|---|
| FHNW retailer-study summary, 10 September 2026 and sample context | Self-reported efficiency benefits and continued human review among surveyed retailers | OpenCart-specific demand, causality, or willingness to pay for our product |
| DHL business study release, 7 October 2025 | Broad multi-market e-commerce business context | A forecast of our adoption or a narrow target-segment estimate |
| Shopify Sidekick help and vendor listing | Direct native merchant-assistant competition and its advertised packaging | Results on a tested merchant workflow |
| WooCommerce MCP developer documentation | Native integration direction and current preview constraints | A complete agent product or stable compatibility across installed shops and plugins |
| WooCommerce for Claude, 7 May 2026 | A vendor experiment beyond bare tool access | General availability or production reliability |
| PrestaShop roadmap | An announced native back-office assistant direction | Shipping status when the entry remains in progress |
| Knowband OpenCart AI Admin Assistant | A directly relevant marketed product and displayed price | Adoption, independent efficacy or absence of unlisted capabilities |
| Metorik pricing/features | Existing paid alternatives for reporting and automation | An appropriate price for Store Deputy |
| Gorgias product site and Gorgias action documentation | Customer-service agent and helpdesk, skills, and operational actions; a primary benchmark for our intended customer agent | Equivalence to owner-led bulk catalog work |
| StoreAgent product, agent offering, MCP and pricing | Packaged WooCommerce competitor covering customer chat, content and managed tool access; published plan limits | Tested reliability, security, OpenCart support or parity with supplier updates and scheduled comment checks |
| n8n WooCommerce integration | A configurable automation substitute | Actual setup burden for our target merchants |
| Adobe commerce standards announcement, 18 February 2026 | The separate discovery/checkout opportunity | A reason to include agent checkout in this MVP |
| OpenCart marketplace and extension guide | A distribution channel and an extension mechanism | Our connector's compatibility or complete admin API coverage |
| Operational AI discussion and cost-reporting discussion | Qualitative language and interview leads, including skepticism | Verified identities, representative demand or measured results |
Conflicts and missing evidence
WooCommerce marketplace plugin documentation and native developer documentation describe different integration surfaces. The native developer page explicitly deprecates the older Woo-specific endpoint; prefer version-specific core documentation for a future adapter. No endpoint is selected for a Store Deputy WooCommerce implementation yet.
The older OpenCart documentation URL /en-gb/system/users/api returned “Page Not Found”. It is not evidence of supported write operations. Inspect the exact installed version during the connector spike.
The DHL press release and the searchable business-report excerpt use different AI-adoption headline wording. We have not reconciled their denominators, so no single AI-adoption percentage from that study is used in this proposal.
Missing: observed OpenCart workflows, merchant interviews, measured task frequency, supplier-file samples, verified platform/version mix, keyword volumes, conversion rates, a reachable-market estimate and willingness to pay. The proposed first workflow is a synthesis to test, not the direct conclusion of any one survey.
Internal architecture evidence
Agentfy's api/src/slices/agent/app/app.prisma and api/src/slices/runtime/README.md were inspected as candidate-reuse evidence. They do not establish a functioning Store Deputy connector. CleanSlice MCP guidance was consulted for dependency direction; it is an architectural standard, not implementation proof.
Recheck vendor capabilities and prices before a competitive pitch or a connector commitment. Replace hypotheses with observed evidence only after recording the environment and method of validation.
Positioning revision
The founder selected Sidekick and Gorgias as the primary product benchmarks and included customer agents in the intended Store Deputy scope. This is a product-direction decision, not evidence of shipped features. MCP is classified as a tool layer; WooCommerce for Claude remains a separate agent/skills experiment. Secondary alternatives are not claimed to be broken or commercially irrelevant. See Market.
StoreAgent was subsequently added to the primary shortlist after the founder identified it. Its bundled customer/content agents must be distinguished from bare WooCommerce MCP. Public pages disagree about renewal of free chat credits: the pricing page explicitly describes Lite credits as one-time. No account or plugin was tested.