Send us the catalog. Save your model for the thinking. Product titles in, structured product data out: types, attributes, missing facts, categories and buyer keywords.
What it is: a low-cost API that turns product titles into structured product data: product type, brand, attributes, missing facts, Google category, collections, search phrases with volume and CPC, and buyer questions, one title per request. Send up to 500 titles at once to group a catalog. 25 tokens per field, with free tokens every day.
Reach Dog tokens are our billing units, not model tokens.
The demo opens huggingface.co. No key and no signup needed.
curl https://api.reach.dog/v1/title-matches \
-H "Authorization: Bearer $REACHDOG_API_KEY" \
-H "Idempotency-Key: $REACHDOG_REQUEST_ID" \
-H "Content-Type: application/json" \
-d '{"title": "Nordic Kayaks NK K1 Kayak",
"parameters": ["product_type", "brand",
"missing_attributes", "search_volume"]}'{
"product_type": "Kayak",
"brand": "Nordic Kayaks",
"missing_attributes": ["Length", "Width", ...],
"search_volume": [
{"text": "kayaks for sale", "search_volume": 40500},
{"text": "kayak sale", "search_volume": 2900}, ...
]
}4 fields, 100 tokens. Values from a recorded example in the title parameter docs.
Good. Keep using it. Just don't make it do this part.
Throwing a raw catalog at a frontier model works, but it is slow and expensive, and you end up re-prompting to fix guessed facts. Reach Dog does the product-data pass first, so your model starts further along.
Keep your frontier tokens for the hard part
Tagging and extracting product facts is grunt work. Do it here at 25 tokens a field, and spend your Claude or GPT budget on the analysis you actually care about.
Skip a couple of passes
The data comes back structured and grounded in product listings that online stores publish, which we collect and process. Attributes come only from the title and any description or attributes you send, and missing_attributes lists what the title leaves out, so you are not re-prompting to clean up.
Get a file you can reuse
Results come back as JSON keyed by your own product IDs. Download it, put it in your project, and feed it to every prompt after that. Replays with the same key are free.
- Send your titlesOne title, or a catalog of up to 500 in one call.
- Get a JSON file backProduct facts, categories, collections, keywords and demand.
- Hand it to your LLMClaude or GPT starts from real data and goes straight to the deep work.
The price is simple, and you see it before you spend.
per field, per title. Pick only the fields you need.
for a catalog output, whether you send 10 titles or 500.
Free tokens every day, so you can try it on your own titles.
Reach Dog tokens are our billing units, not model tokens. They are separate from what you pay Anthropic, OpenAI or any other model provider.
- Catalog jobs show you a quote and wait for your OK before anything runs.
- Repeating a request with the same key returns the saved answer at no charge.
68 fields. Ask only for the ones you need.
Every request lists the fields it wants, and you pay for those fields only. Here are some of them.
Title38 fields
POST /v1/title-matches What a product is and how people search for it.
product_typebrandattributesmissing_attributescategory_pathscollection_suggestionsfacetssearch_phrasesbuyer_questionssearch_volumecpcrelated_product_types
Research14 fields
POST /v1/market-research The market around a product type.
demand_trendgeographic_demandmarket_coveragebrand_presenceseller_concentrationcomparable_listings
Content10 fields
POST /v1/content-generations Content written from a result you already have.
content_briefcontent_draftsad_topicsblog_topicsvideo_topicsfirst_test
Catalog
POST /v1/batches Your whole catalog at once, up to 500 titles.
catalog_summarycatalog_groupsgroup_membersungrouped_itemscatalog_keyword_opportunitiescatalog_content_ideas
10 ready-made workflows to start from.
Each recipe combines the right requests for a job. You load your titles, see the full request plan, and nothing is charged until you run it.
- 01
Demand-space lenses
Turn bare titles into overlapping product groups, each with the reason its products are in it.
- 02
SEM ad campaign deck
Keywords with volume and CPC, ad angles from stated facts, and a first test per product.
- 03
Social and video launch kit
A 30-day content calendar across channels, each slot tied to its product.
- 04
Shopify and Algolia proposals
Collection memberships and search records mapped to your IDs. Export only.
- 05
Market whitespace memo
Keyword demand set against how many listings already chase each term, with every line linked to its evidence.
- 06
Shopping assistant grounding
One evidence document per product, plus instructions so your assistant answers only from it.
- 07
Cross-border expansion
Local search terms and local demand for new countries, kept apart.
- 08
Cart bundles and cross-sell
Product pairs from your own catalog, with the reason for each pair.
- 09
Supplier feed auditor
Find missing fields and write the supplier questions. It never fills in values.
- 10
Seasonal calendar and AEO
Monthly demand next to weekly supply, plus the questions buyers ask, ready for a buying guide.
Use it from the tools you already have.
REST API
Call it from any language. Python and JavaScript quickstarts are ready to copy.
POST api.reach.dog/v1/title-matches MCP server
Ask Claude Desktop or any MCP client to look products up for you.
npx --yes @reachdog/mcp@0.1.0 LangChain
Give your agent product tools with a budget and an approval step.
pip install langchain-reachdog==0.1.0 n8n
Audit a product feed from a Google Sheet, with an approval stop before it spends.
n8n.io/workflows/22575 Hugging Face
See real recorded requests and responses. No key, no signup, no tokens spent.
ReachDog/product-classification-api-demo Built from the market, not from a model's guess.
Source: Reach Dog's commerce map, built from product listings that online stores publish, which we collect and process, plus keyword data. Counts last reviewed November 2026. Keyword metric fields such as search volume and CPC state their own market, period and provider.