New Research with Claude · Query millions of auction results from your own AI toolsGet Started
Guide · Collectors

How collectors research a market and time a buy or sell

By ALT/FNDATA · October 2026 · 6 min read

Two questions sit behind every collecting decision: what is this really worth, and is now the moment to buy or to sell? Asking prices and forum opinion will not answer either. Realized auction results will, and the connector puts them inside the AI assistant you already use, with a link to the sale on every figure so you can check it yourself.

Get your API keyConnect your assistant
YouHermès Birkin 30 price in 2024?AALT/FNDATAIllustrative$12,000 – $38,000Realized salesChristie's · 2024$20,160↗Sotheby's · 2024$17,500↗Bonhams · 2024$12,600↗Every result links to the sale · auction house · date
The aha moment: ask in plain language, get a sourced answer
In short

Chain three tools: search to see what recent examples fetched, price_trend to read the direction, and price_check before you bid or set a reserve. Read realized, not asking. Every result links to the sale, so you are never trusting a bare number.

01How it works

Realized prices, inside your own assistant

The connector adds auction price data to Claude, ChatGPT, Gemini and other MCP-capable assistants. You ask in plain language across 20 categories from 650+ auction houses, the assistant calls the right read-only tool, and each answer comes back normalized to USD with the auction house, the sale date and a link.

02The workflow

Research a piece, then time the decision

See what comparable examples sold for, read the trend, and run a check before you act.

1

See what recent examples fetched

Ask your assistant“What did a Hermès Birkin 30 sell for in 2024?”
Toolssearchmarket_stats
What comes back

Matching sold lots merged by realized price, or a statistical summary of the segment: the median and the spread, with the leading houses. You learn the going rate from real sales, not from a listing someone hopes to achieve.

Tip. Add the detail that moves price (size, material, colour, hardware). “Birkin 30 Togo black” is a far sharper question than “Birkin”.
2

Read the direction of the market

Ask your assistant“How have Hermès Birkin prices moved over the last five years?”
Toolsprice_trend
What comes back

Realized prices bucketed by year or quarter, with the count and median per period. This is what answers “is this still climbing or has it cooled?” before you commit either way.

Tip. A rising median on healthy volume is a different signal from a rising median on only a few sales. Watch the count alongside the median.
3

Check fair value before you act

Ask your assistant“Price-check a Louis Vuitton OnTheGo PM in black Empreinte leather.”
Toolsprice_check
What comes back

A fair-value range for that item from comparable sales, with the comparables and a confidence label. A well-specified piece reads high; use the range to set a ceiling for a bid or a floor for a reserve.

Tip. Specify the piece fully. A detailed request reads high confidence; a vague one reads low and flagged, which is your cue to add detail.
4

Explore across the whole market

Ask your assistant“Search the whole market for a Burma ruby ring over $50,000.”
Toolssearch
What comes back

A keyword search across every category at once, merged by realized price, for whatever you are curious about. A good way to learn a corner of the market before you spend in it.

Tip. Treat search as the front door: from a result, ask the assistant to run a trend or a price check on the closest match.
03Reading the answer

Confidence, caveats and why the number is defensible

Price Check labels its answer high, medium or low, and it is honest when the evidence is thin. A specific, well-described item with a tight band of recent sales reads high. A vague request reads low and is flagged as a loose match, and a genuinely wide price spread is flagged and capped down. If you get a low-confidence range, add detail and ask again. The point for a collector is that you are never asked to trust a bare figure: every result carries the sale behind it, so you can open the source and judge for yourself whether the comparables really match your piece.
04Honest limits

What the data does and does not cover

A few things to keep in mind as you research.

  • It reports realized auction results, what pieces actually sold for, not retail or private-sale asking prices.
  • Single-item fair-value checks cover handbags, watches and cars. For jewelry, gems, art and other categories, use search and market_stats to read the market.
  • Colored-stone single-item comparables are held back, so for a gemstone piece lean on market statistics.
  • Condition and specification drive price, so read the detail on a result before treating it as a match for your piece.
  • It is research, not personalized investment advice. A collecting decision is still yours to make.
05Getting started

Connect it once, then ask in plain language

Access requires a paid ALT/FNDATA API key. Get one in the app, add the connector to the assistant you use with the setup page for your tool, and start asking.

1Get an API keyPaid access in the ALT/FNDATA app. There is no free tier.2Add the connectorFollow the setup page for Claude, ChatGPT, Gemini, Perplexity or your tool.
3Ask awayUse the prompts above. Every result carries a source link you can cite.

Questions

Frequently asked questions

Can it tell me whether to buy or sell now?
It gives you the evidence to decide: what comparable pieces recently fetched, how the market has moved, and a fair-value range for your item. The decision stays yours; this is research, not personalized advice.
Does it show asking prices or realized sales?
Realized auction results, normalized to USD, each with the auction house and sale date. That is what an item actually fetched, not what a seller hoped for.
Which assistants can I use it in?
Any assistant that supports MCP connectors, including Claude, ChatGPT, Gemini and Perplexity. The setup pages cover each one.
How do I get better answers?
Be specific. Name the model, size, material, colour or reference. A detailed question returns a tighter comparable set and a higher-confidence read.
Is it free?
No. Access requires a paid ALT/FNDATA API key, which you create in the app.
Keep reading

Related guides

Valuing client collectiblesWatch comps and price trendsJewelry appraisal comparablesAll datasets

Put a sourced number on it today.

Connect the ALT/FNDATA connector to the AI assistant you already use, and pull realized auction comparables, market statistics and price trends in plain language, each with a link to the sale.

Get your API keySee the setup pages

The ALT/FNDATA connector is read-only and returns realized auction results across 20 categories from 650+ auction houses, normalized to USD, each with the auction house, sale date and a source link. It reports what items sold for at auction, not retail or private-treaty prices. Methodology · All datasets · All guides