@syntalic/mcp-server; source and release history) exposes 17 tools by default: 6 free and 11 paid measure tools, one per measure. Below is what each does and what to ask in natural language. Claude (or your MCP client) fills in the arguments; you don’t have to think in parameters.
Start with the free Discovery tools. They cost nothing and tell you whether a paid query will land on good data before you spend on it.
Upgrading from 0.12 or earlier? The per-route tools (
best_price, price_positioning, ontology_resolve and the rest) are now splits of the measure tools. To get them back as separate tools, set SYNTALIC_TOOLS=routes. See Per-route tools.Discovery
Free — no payment required.catalog_overview
The shape of the catalog: unique products, listings, brands, categories, retailers, price observations, and when it was last updated.
“What’s in the Syntalic catalog?” “How many products and retailers do you actually cover?”
coverage_map
Where the catalog is deep vs. thin, as priced product counts and a quality status per retailer × country × category. The one to check before spending.
“Do you have good coverage of food and beverages in Canada?” “Which categories are thin on Best Buy?”
browse_categories
The product category tree with listing counts per node (one per store, variant and country). Pass a parent path to drill into a subtree.
“What categories do you cover?” “Show me what’s under Home & Garden.”
list_retailers
Retailers in the catalog with listing counts, countries covered, and freshness.
“Which retailers do you cover in Canada?” “How fresh is your Costco data?”
list_brands
Brands in the catalog with listing counts. An optional query prefix-matches the name, so “sam” finds Samsung. To see the brands on a shelf, use measure_assortment.
“Do you have Anker products?” “Which brands do you have that start with Sam?”
wallet_info
Your wallet addresses, the balance on each chain, and funding instructions.
“What’s my Syntalic wallet balance and where do I send USDC?”
Measure tools
Each tool acceptssubject (an id from measure_resolve, or a name where supported) and by (the split; each tool has a default). Both are optional in the tool schema; the selected split determines which route-specific parameters are required. Do not send a subject and its equivalent route parameter together. Each split is answered by one API endpoint, named in the table, and costs what that endpoint costs: the MCP calls the endpoint itself, so a split on a $0.01 endpoint is $0.01 even when the measure’s own HTTP path is $0.02.
An unsupported split or incompatible subject is rejected before the request is sent. Resolve names to IDs when possible; ambiguous names can return multiple co-equal matches, so select the appropriate kind rather than assuming the first match is correct.
measure_offers
$0.01. Who sells one product, and at what price, now.
“What’s the cheapest place to buy AirPods Pro right now?” “Find me the best price for a Dyson V15 in Canada.”
measure_prices
$0.01 to $0.02. The price level and spread of a shelf.
“Give me a pricing summary for cordless vacuums.” “Where does Anker sit on the wireless charging price ladder?” “Is Costco cheaper than the grocery shelf on average?”
measure_trend
$0.01 to $0.02. How prices moved, for a product, a shelf or a brand.
period is about a product or a brand, never a shelf. For a shelf, ask for the biggest movers (by: product).
“How has the Sony WH-1000XM5 changed in price over the last 90 days?” “Which electronics products moved the most in the last 30 days?”
measure_promotions
$0.01. Who discounts, how often and how deep.
“Which retailers discount grocery most often?” “Show me the best deals on espresso machines this week.”
measure_assortment
$0.01 to $0.02. What is on the shelf, and who holds it.
“What’s Apple’s share of shelf in wireless earbuds?” “Which retailers carry Olipop?”
measure_attention
$0.03. Share of social conversation (TikTok and Instagram), and how it is moving.
“What’s the share of conversation by brand in food-beverages-tobacco?” “Which brands are rising in health-beauty?” “Chart weekly mentions for Olipop.”
measure_signals
$0.03. What posts say, as labels: launches, reviews, stance, names, practices.
With
by: practice, the channel filter takes the nine channels tactics takes: email, retail, supply, brand_launch, growth, social, site, branding and customer. Leave it out to see practices in every channel. Before MCP server 0.14.0 it took only email.
“What are creators saying about new soda launches?” “Which marketing practices do creators say beauty brands use on email?”
measure_scout
$0.05. Social attention set against the shelf: the only tool that binds both.
“Which soda brands have more conversation than shelf share?” “Which new snack listings landed with social traction?”
measure_resolve
$0.01. A name, phrase or code to ids. Call it first when a question names a brand, shelf, product type or retailer, and scope the other tools with the ids it returns.
“What does ‘protein bars’ map to in Syntalic?” “What attributes does GS1 define for coffee?”
measure_neighbors
$0.01. What sits next to an id: ancestors, children, product types, brands, and the chains that carry it (by: relations, the only split).
For a shelf (spt:) it also answers two social blocks:
facets, in the defaults: the marketing practices creators describe brands using on the shelf, grouped by channel (email, retail, supply, brand launch, growth, social, site, branding and customer care), with the creators and posts behind each.labels, only when you ask for it by name (relations: "labels"(orrelations: ["labels"]in per-route mode)): what the posts about the shelf are. Each post carries one to three kinds (advice, story, pitch, showcase, demo, news, insight, review), up to three themes (what it is about) and facets such asfor_operators,compares_brandsorpre_launch. Every item hascreators,posts,brandsand up to fiveexample_brands, over the last 30 days. Read the counts as the tracked corpus (TikTok and Instagram), not as a share of the market.status: unavailablemeans the labels couldn’t be read for this answer, either because they aren’t published yet or because of a temporary problem. It doesn’t mean there are none.
“Which chains carry this brand, and which don’t?” “What kinds of posts are about soda, and who are they aimed at?”
measure_coverage
$0.01. How much evidence stands behind an id: listings, freshness, channels, and the chains scanned (by: evidence, the only split). Use it before saying a brand is “not seen” at a retailer: with a brand id and a retailer, it says whether the brand was seen there, the retailer was checked and the brand not seen, or the retailer wasn’t checked.
“How much data do you have on Olipop, and which retailers did you check?”
Per-route tools
SetSYNTALIC_TOOLS=routes to replace the 11 measure tools with per-route tools: 39 tools, the same 6 free ones and 33 paid, one per endpoint except the five listed below. This is the 0.12 tool set without the inflation tool, which is currently unavailable. The playbook prompts still name these tools. In default mode, use the corresponding measure_* tool and split rather than a bare route or measure name.
Five endpoints have measure splits but no standalone MCP tool (they are still callable over HTTP):
product-share, brand-signals, emerging-brands, product-names and tactics. The inflation and similar-brands endpoints are currently unavailable, so the MCP doesn’t offer them. Shelf movers are available via measure_trend with by: product; this is not an inflation statistic.
How to set the variable in your MCP config is on the MCP server page.
Filters
Besidessubject and by, tools accept route-specific parameters. An MCP client can infer them from a prompt, or you can supply them explicitly. The selected split determines which are valid; check the tool schema and API Reference.
Social and Scout splits accept the following where supported (not all splits accept every filter):
For
attention by emerging and signals by product_name, use an spt: ID or the split’s explicit filters, not a plain category name as subject.
A 404 NOT_PUBLISHED on a social split means that view isn’t published in the current data release, not that the shelf is quiet. Nothing is billed for it.
Sample response
Illustrative, abridgedmeasure_offers response (not a live quote):
Prompts and resources
The MCP also serves playbook prompts and resources (syntalic://playbooks, syntalic://ontology/rules, and syntalic://ontology/metrics). The playbook prompts and the rules and metrics resources name the per-route tools, such as price_positioning or promo_intelligence, and some mention the unavailable inflation tool. In the default tool set, use the matching measure tool and split instead (price_positioning is measure_prices by brand), and skip inflation steps. See the MCP server repository.