How ALT/FNDATA works in Claude
ALT/FNDATA connects to Claude through the Model Context Protocol, so you can ask about verified auction and private-sale prices in plain language and get answers built from real transaction rows. Here is what happens between your question and the answer.
What is MCP
The Model Context Protocol (MCP) is an open standard that lets Claude call an external data source live during a conversation. Instead of pasting a spreadsheet into the chat, you connect ALT/FNDATA once and Claude queries it whenever a question needs real numbers. It reads the tool descriptions, decides which to call, fills in the parameters, and folds the result into its answer.
The three-step flow
Every answer follows the same path, whether you are asking for one comparable or a five-year trend.
-
You ask in plain language
Name the data ("Using ALT/FNDATA"), a category, and a period. No SQL and no export step. For example: "Using ALT/FNDATA, show the ten highest Patek Philippe watch prices sold in 2024."
-
Claude queries the connector
Claude reads the table schema so it uses real column names, then runs one read-only query with the right filters and sort on your behalf.
-
You get a grounded answer
The reply is built from verified secondary-market and private-sale rows, not model guesses, and you can see every query Claude ran.
Read-only and auditable
The connector exposes three read-only tools. Nothing is ever written or changed, and every query Claude runs is visible to you, so you can check exactly which table, filters, and sort produced an answer.
Connect the connector
Getting ALT/FNDATA into Claude takes three steps: get an API key from the portal, add the server URL https://mcp.altfndata.com/mcp as a custom connector, and ask your first question. It works in Claude on the web and desktop through the remote connector, or locally for Claude Desktop and Claude Code.
Follow the full Quickstart for step-by-step setup across all three clients, with the exact config and commands.
The three tools
The connector registers three tools. All are read-only and return Markdown-formatted text that Claude reads back to you. Column names are validated live against each table's real schema, so Claude starts with get_table_schema before filtering on a field it is unsure of. You rarely call these yourself; you ask in plain language and Claude picks the tool.
| Tool | What it does | Key parameters |
|---|---|---|
list_tables | Lists the tables your key can reach, with a one-line description and column count for each. | None |
get_table_schema | Returns the column names and types for one table, so filters and sorts use fields that exist. | table |
query_table | The workhorse. Filters, sorts, selects fields, and paginates, then returns matching rows. | table, fields, filters, sort_field, sort_direction, limit, offset |
Filter operators
Each filter is {"field": ..., "op": ..., "value": ...}. Available operators:
| Operator | Meaning | Example value |
|---|---|---|
eq / neq | Equals / not equals | "Rolex" |
gt / gte | Greater than / or equal | 50000 |
lt / lte | Less than / or equal | 1000 |
like | Contains (substring match) | "Daytona" |
in / not_in | In / not in a list | ["Christie's","Sotheby's"] |
is_null / is_not_null | Field is empty / present | (no value) |
offset, or use the underlying REST API, which supports aggregations and group_by directly.What you can reach
Every category shares one wide schema, so the same fields and the same query work across tables. Swap the table alias to move a question from watches to handbags to fine art. The backend registers these aliases; which ones your key can query depends on your plan, so ask Claude to run list_tables to see your set.
| Alias | Coverage |
|---|---|
watches | Luxury watch auction results |
handbags | Luxury handbag auction results |
jewelry | Jewelry and gemstone auction results |
gems | Loose gemstone auction results |
automobile | Collector car auction results |
motorcycle | Collector motorcycle auction results |
aircraft | Aircraft auction results |
wine_whisky | Wine and whisky auction results |
works_of_art | Works of art with stone or jewelry relevance |
fine_art | Fine art across movements and periods |
The connector's built-in guidance highlights watches, handbags, and jewelry as the primary set; the other aliases are available where your plan includes them.
Documented fields
These are the fields we document and support across the category tables. Ask Claude to read the full column list for any table with get_table_schema.
| Field | What it holds | Use it for |
|---|---|---|
designer | Brand or maker, for example Rolex or Hermes | The main brand filter |
model | Model line, for example Daytona | Narrowing within a brand |
item_title | Full lot title and description | Free-text matching with like |
sale_date | Date the lot was offered or sold | Any time-based question |
usd_price_decimal | Realized price in USD | Prices; populated when status = sold |
sale_estimates_high_usd_price | Pre-sale high estimate in USD | A forecast, not a paid price |
status | sold or unsold | Filter to sold before reading a price |
vendor | Auction house or marketplace | Grouping and source checks |
stock_ticker | Ticker of the brand's public parent, for example CFR.SW | Linking a brand to a listed company |
Sample prompts
Copy any of these into Claude once the connector is on. Each names the data and a category, and each answer carries the caveat noted below it. The follow-ups show how to drill in without starting over. There is a much fuller set, eight roles and more than twenty recipes, in the use-case playbook.
Collection allocation & liquidity
Family officeUnderstand what a family's collecting categories are worth at the market level, and how liquid each is, without pretending recent quarters are complete.
Carry: newest quarters undercount; the most recent year is provisional.
Luxury demand read-through to equities
Hedge-fund analystUse auction demand as an alternative signal on a listed brand's health, and tie it back to the parent company's ticker.
Carry: secondary-market demand, not primary sales or revenue; recent quarters provisional.
Client holdings valuation context
Wealth advisorGive a client a defensible market range for an asset they hold or want, grounded in comparable sold lots rather than a single data point.
Carry: a market range from comparables, not an appraisal; excludes retail and dealer asking prices.
Competitive resale benchmarking
Luxury-brand strategistBenchmark a house's resale strength against rivals: how prices hold versus estimate, and where demand is trending across categories.
Carry: auction resale behavior, not primary retail performance; recent periods provisional.
Comparable-sales evidence
AppraiserAssemble a clean, sourced set of comparable sold lots for a valuation file, with the auction house and date on every comparable.
Carry: comparables from secondary-market sales; verify condition and specifics against each lot.
Market-trend fact-checking
Financial journalistStand up or knock down a market-trend claim with sold-lot evidence, and avoid mistaking ingestion lag for a real move.
Carry: a dip in the last one to two quarters is likely ingestion lag, not a market decline.
Limits & caveats
The data is honest about what it is. Carry these when you read an answer, and see the coverage and methodology page for the full treatment.
- Secondary market plus private sales. Coverage is auction results and private-sale transactions, not retail list prices or dealer asking prices.
- The newest one to two quarters undercount. Recent sales are still being ingested, so the latest periods look thinner than they are. Treat very recent volume and counts as provisional, not as a market decline.
- Read prices only on sold lots.
usd_price_decimalis populated whenstatus = sold. Estimates are forecasts, not amounts paid. - Prices are USD-normalized using nearest-date exchange rates, so a converted figure can differ slightly from a same-day conversion.
- Rows, not aggregates, over the wire. Any average or median Claude reports is computed over the returned rows, up to 500 per call. For population-level statistics, page through results or use the REST API.
- Not investment advice. The data is for research and reference.
Where to next
- Quickstart. Get a key and connect the connector in Claude web, Desktop, or Claude Code.
- Use-case playbook. Eight roles and more than twenty copy-paste recipes, each with its caveat.
- Technical docs. The three tools in full, the dataset and field reference, auth, and limits.
- Coverage and methodology. What the data covers, how fresh it is, and how to trust an answer.