TikTok Shop product research,
callable from your agent
Most Shop tools hand you a dashboard and leave the decision to you. HookLayer returns the decision itself, as structured output your code or your agent can act on. Scan a niche, get a ranked shortlist, and get a verdict per product with the reason it was reached.
Two tools, one scope
Discovery and depth are separate calls so you only pay for the depth you actually want.
product_scoutcosts 5 creditsScans a niche and returns ranked products with provider evidence and a PROMOTE, WATCH or SKIP verdict per row. The ranking path is deterministic scoring with no model in the loop, so the same inputs produce the same order.
analyze_productcosts 10 creditsGoes deep on a single listing: the buyer proof behind the rating, the recurring complaints in the reviews, and whether those objections are answerable on camera. Those complaints are the script.
Every product comes back judged
A verdict without a reason is just an opinion, so the reason travels with it in the response.
Buyer proof is real and the price clears the floor where a video can repay the hours it costs.
Something promising is missing a signal, most often review volume that is too thin to trust yet.
A disqualifying signal was found. The verdict names it rather than leaving you to infer it.
What the response will never contain
TikTok does not publish per-product revenue. Any tool that shows one is deriving it, normally by multiplying a sold count by a price, and presenting a derivation as a measurement is the single most common way this category misleads people. HookLayer returns a field called Sales Proxy, which is exactly that derivation, labelled as a proxy and documented as not actual or historical revenue. It is there to rank one row against another. It is not a forecast, and it is not presented as one.
Call it directly
The MCP tool and the REST endpoint run the same implementation, so an agent and a scheduled job share one integration.
curl -X POST https://hooklayer.dev/api/v1/product/scout \
-H "Authorization: Bearer $HOOKLAYER_API_KEY" \
-H "Content-Type: application/json" \
-d '{"niche": "beauty", "region": "US"}'Questions
How do I call TikTok Shop product research from Claude or ChatGPT?
Connect HookLayer as an MCP server, then ask in plain language. The host calls product_scout with your niche and region, and the tool returns ranked products, each with a PROMOTE, WATCH or SKIP verdict and the sentence explaining which signal decided it. Because HookLayer exposes all of its tools on one surface, the same conversation can continue into score_hook or viral_remix to write the script from the buyer language in the reviews. No separate account or key is needed beyond your HookLayer credentials.
What is the difference between product_scout and analyze_product?
product_scout is the discovery call: give it a niche and it scans, ranks and returns a shortlist with a verdict per product. It costs 5 credits. analyze_product is the depth call: give it one product and it returns the full breakdown for that listing, including the review-derived objections you would answer on camera. It costs 10 credits. The usual pattern is one scout call to find candidates, then analyze_product on the one or two worth filming.
Does HookLayer return TikTok Shop revenue or GMV?
No. TikTok does not publish per-product revenue, so any tool showing a revenue figure is deriving it, normally by multiplying a sold count by a price. HookLayer does not present a derived number as measured. The response carries a field called Sales Proxy, which is the provider-reported cumulative sold count multiplied by the current price. It is labelled a proxy, it is documented as not actual or historical revenue, and it is there for ranking one row against another rather than as a currency forecast.
Which plans include the Shop tools?
product_scout and analyze_product are available on Pro and Agency, not on Free or Creator. The reason is cost rather than packaging. Every scan makes live calls to a commerce data provider, so unlike a purely prompt-based tool its cost scales directly with usage. Both tools share a single product_scout permission scope, so an API key authorised for one can call the other.
Can I call the Shop tools over plain REST instead of MCP?
Yes. Every HookLayer tool is exposed as both an MCP tool and a REST endpoint against the same implementation, so there is no behavioural drift between the two. Authenticate with your HookLayer API key as a bearer token and POST to the endpoint. The MCP surface is the same contract with a different envelope, which is what lets an agent and a cron job share one integration.
Connect it once, call it from anywhere
HookLayer runs as an MCP server and a REST API over the same 14 tools, so Claude, ChatGPT, Cursor and your own scripts all hit one surface.
Setup instructions