SuggestAPI for Claude Commerce

Summary. Give Claude Commerce agents better product discovery from your catalog. Try the live demo, or clone the GitHub adapter that implements search_products against SuggestAPI.

Keep the search engine, catalog, and checkout. Add a discovery layer so agents can find, compare, recommend, and hand off without dropping the shopper’s intent. A live demo and GitHub adapter let you try that loop yourself.

Anthropic’s Claude for commerce blueprint is the conversation layer. Its tools should call the merchant systems already in place, and product search should return results that are already ranked. [1] [2] SuggestAPI is the discovery adapter in that pattern, not an Anthropic product. Cart, checkout, fulfillment, and payment stay with the merchant.

Claude holds the conversation; your systems stay the source of truth

A shopper rarely arrives with a tidy keyword. They ask for waterproof hiking shoes under $180, then add a delivery window. Claude for commerce holds that conversation. SuggestAPI turns the job into current, ranked products from the catalog you already run.

shopper job → Claude → SuggestAPI → your catalog → shortlist → your checkout

Claude reasons about the goal. SuggestAPI finds the products. Your systems stay the source of truth.

You don’t need a new discovery stack for every agent

Search, ranking, inventory, and checkout already encode years of merchant work. An agent should use that, not imitate it with a prompt. SuggestAPI sits in front: it holds the shopping job through refinements, retrieves against Algolia, Typesense, Elasticsearch, Meilisearch, Shopify, or a SuggestAPI index, and returns product IDs, prices, availability, and attributes the agent can present without inventing catalog facts. [3] [4]

There’s no search migration and no second catalog. “Something warm for Iceland in October under $250” stays a job, not a reset keyword.

Build search_products once

Put SuggestAPI behind the agent’s product-search tool. Start with search and product details; grow into compare and recommend later. [5]

search_products(job, filters, price_range, limit) → SuggestAPI → ranked products

Try it yourself

Claude Commerce already calls search_products and get_product_details. We published a starter StorefrontBackend that implements those methods against SuggestAPI, plus a shopper you can run in the browser.

Open the live demo. Tap a chat example, add a result to the cart, then hand off to merchant checkout. The tool log is the same search_products → add_to_cart → checkout contract Anthropic’s shopping agent uses. The public instance may take about a minute to wake.

Clone suggestapi/suggestapi-claude-commerce-agents to run it locally. Python 3.11+. No extra packages, no API keys for the guest catalog.

python3 tests/test_harness.py
python3 demo/server.py

The harness prints each tool call and the ranked products. The browser demo does not call Claude; it shows the contract. Point Anthropic’s shopping agent at SuggestAPIStorefront when you want the model in the loop. Checkout stays on the merchant storefront. SuggestAPI does not take payment. [7] [8]

A clean division of work

Layer	Job
Claude for commerce	Conversation, comparison, recommendation
SuggestAPI	Intent, retrieval, ranking, structured product context
Merchant systems	Catalog, inventory, cart, checkout

The same discovery layer can serve the storefront and the agent. The job should not disappear when the surface changes. [6]

Frequently asked questions

Does SuggestAPI replace Claude for commerce?

No. Claude is the conversation layer. SuggestAPI is the product-discovery layer behind search and retrieval.

Do we need to replace our current search provider?

No. SuggestAPI sits in front of the search and catalog you already run.

What does the agent receive from SuggestAPI?

Current, ranked product candidates and structured fields for compare, recommend, and present. Product IDs connect to cart and checkout.

Can this support follow-up questions?

Yes. The shopping job stays available as the customer adds constraints or moves from finding to comparing.

Is SuggestAPI an Anthropic product or official Claude integration?

No. It is an independent adapter for a commerce agent built on Claude, not an Anthropic product or certified component of the blueprint.

Can we try this without wiring Claude?

Yes. The live demo and local harness use a guest catalog and the same search_products, add_to_cart, and checkout_handoff tools. Claude is not required until you point Anthropic’s shopping agent at SuggestAPIStorefront.

Power product discovery for Claude Commerce agents