Few searches products on Shopify and tells you how good each one looks. A REST API and an MCP server, backed by a visual taste model.

tote-api

Tote searches its own Shopify product index and judges visual taste from product photos and store judgments. It estimates aesthetic preference; it does not verify materials, construction or durability.

Hosted REST search and verdict are available. The repository provides JavaScript exports for both, and the hosted MCP endpoint exposes both tools. npm publication and named MCP client validation remain pending. Use REST now, or install the JavaScript wrapper from a local checkout. MCP client integrations remain a preview; protocol-level validation does not establish compatibility with every client.

The working service hostname is https://tote-api-4e6mvag77q-uc.a.run.app. api.tote.store is not verified and is not used in examples. Laurent retains the npm testing/publication step.

Keep your own search and send Tote what you found: if Tote is unavailable, your results are unchanged. Search runs on a single instance today. Expect a few seconds on a fresh query, under a second on a repeated one, and plan for roughly ten fresh searches a minute; these are the September 14, 2026 measurements in the capacity report, not a service level. Ask for a key before building on it.

tote_demo has 200 units per UTC day, shared across all users, resetting at midnight UTC. Each search request consumes one unit; each submitted verdict product consumes one unit. Accepted failures and repeated calls consume units. The website uses its own server-held key. For a separate key, message me.