Track Polymarket Whales and Top Traders with Claude

A whale tracker answers three questions. Who is trading the most? Are they betting or just making markets? And is a given trader winning? This page walks through all three with the Bitquery MCP server in Claude, ChatGPT or Cursor, using six of its 14 Polymarket tools and the address-label tool.

It reads data only. It cannot place or copy a trade. What it gives you is the research step before you decide whether anyone is worth following.

claude mcp add --transport http bitquery https://mcp.bitquery.io

A 38-second demo, Polymarket and Hyperliquid data in ChatGPT and Claude, shows the server at work in a chat.

Every figure below came from the production server on 21 September 2026, between 12:22 and 12:27 UTC, with data 13 to 16 minutes behind the clock.

Step 1: Rank traders by what they traded

"Who were the top Polymarket traders by volume over the last 7 days?"

The agent calls polymarket_top_traders with a 168-hour window.

#	Address	7-day volume	Orders	Orders that took liquidity	Outcomes traded
1	0x2005d16a...75ea	$24,961,757	91,987	14.8%	11,142
2	0xfe787d2d...0319	$24,249,729	63,289	7.0%	5,989
3	0x2c335066...0563	$21,549,928	7,243	10.3%	551
4	0x821dab05...cf16	$20,905,918	152,003	79.6%	29,675
5	0x2a696600...d2f7	$18,645,208	144,594	100%	28,702
6	0x52685279...135d	$16,084,516	31,618	33.3%	4,423
7	0x224a89db...a3d5	$14,817,566	17,551	14.6%	917
8	0x63cf8544...e4d0	$12,104,594	303	11.2%	6

This is a volume ranking. The tool computes no profit here and says so in its first row. The two right-hand columns already sort the list into types. Seven of them placed thousands of orders across hundreds or thousands of outcomes. Number 8 reached $12.1 million with 303 orders on six.

Step 2: Tell a market maker from a bettor

"Give me the lifetime summary of number 1 and number 8."

polymarket_trader_summary returns one row per address, covering its whole recorded history.

	Number 1	Number 8
First trade	1 September 2025	16 September 2026
Orders	4,606,941	303
Orders resting on the book	92.5%	88.8%
Markets traded	141,322	6
Lifetime volume	$597,996,502	$12,104,594

Number 1 has quoted 141,322 markets since the first day of recorded history. That is the footprint of a market maker, which quotes both sides for the spread, so its trades say little about who will win. Number 8 appeared five days before this page was written, bet on six sports results in two days and stopped. That is what people mean by a whale.

Step 3: Read what the whale bet on

"What did 0x63cf8544631c31e9fcea0b02141dd24d01d6e4d0 bet on this week?"

polymarket_trader_activity breaks its orders out by market and outcome.

Market	Bought	Spent	Average price
Lions vs. Bills	Bills	$3,329,463	0.70
Will Real Betis win on 17 September?	Yes	$2,130,921	0.65
Will AC Milan win on 16 September?	Yes	$1,975,524	0.45
Will Chelsea FC win on 18 September?	Yes	$1,070,173	0.38
Will Villarreal CF win on 17 September?	Yes	$90,000	0.45
Will Real Sociedad win on 17 September?	Yes	$36,660	0.33

A price is dollars per share between 0 and 1, so 0.70 means it paid as if Bills had a 70% chance.

Step 4: Score it

"Show its positions, open and closed, with profit and loss."

polymarket_trader_positions with status set to all adds how each market resolved and, where it can, the profit.

Market	Result	Profit or loss (tool)
Real Betis	Won, sold at 0.999	+$1,136,626
Villarreal	Won, sold at 0.999	+$109,800
Real Sociedad	Lost, sold at 0.04	-$32,205
Chelsea	Lost, still holds 2.8 million worthless shares	-$1,070,216
Bills	Won	not computed
AC Milan	Lost	not computed

On the last two rows the tool leaves profit empty and explains why: the shares left the address without a trade, so no sale price exists. It reports the share count as exact and will not guess the rest. The Bills shares left at 03:46:30 UTC on 18 September. polymarket_settlements shows other Bills holders being paid out through a relayer contract in that same second, which fits a payout, though no record ties this address to one.

The arithmetic from here is ours, not the tool's. The four scored markets sum to +$144,005. The Bills shares, 4,726,883 of them, paid $1 each against $3,329,463 spent, a gain of $1,397,421 if they were cashed at the payout, which is our assumption. The AC Milan shares paid nothing, a loss of $1,975,524. Across six markets that is about -$434,000, before $149,358 of trading fees.

So the eighth-largest trader of the week won three of six bets and still lost money. Volume made it look worth following. The positions say otherwise.

Step 5: Watch the fills

"Show its orders in Lions vs. Bills."

polymarket_market_trades with the trader filter lists each order.

Time (UTC, 17 September)	Price	Size
20:28:52	0.71	$698,710
20:28:54	0.69	$56,484
20:28:54	0.70	$1,989,229
20:28:54	0.70	$356,534
20:30:04	0.70	$32,106
21:29:45	0.70	$196,399

$3.1 million went in across four orders inside two seconds, all of them taking liquidity from the book. Drop the trader filter and set a size floor instead, and the same tool becomes a live feed of large fills for any market.

To put a name on an address, pass the list to labels_for_addresses. For four of the addresses above it returned a label on three: two carry a Polymarket account name and number 1 is tagged only as a Polymarket proxy wallet. Number 8 has no label.

A prompt you can reuse

"Rank Polymarket wallets by 7-day volume. Drop any with more than 1,000 orders or more than 100 outcomes traded. For the rest, show the lifetime summary, then open and closed positions with profit and loss, and tell me which markets they hold now."

The order and outcome cut-offs are a starting point taken from the table in step 1. Tune them to the market you follow.

What to know before you rely on it

• Read-only. No order placement and no copy trading. To place orders use Polymarket's own CLOB API.
• Rankings are by volume. Profit lives in polymarket_trader_positions, one address at a time.
• Profit is sometimes empty. When shares arrive or leave without a trade the tool cannot price them and says so. Its own notes put this at roughly a quarter of open positions.
• Profit is before fees. The fee total comes back as its own column.
• One wallet is not one person. A trader can split bets across wallets, and these tools do not link them.
• A few minutes behind. Calls on this page trailed the clock by 13 to 16 minutes.
• History starts in September 2025. A first trade on 1 September 2025 may be older than it looks.

Related

• Polymarket MCP server: all 14 tools, with odds, price history and disputes.
• Polymarket GraphQL API: the same data in code, with WebSocket and Kafka streams, for a tracker that runs by itself.
• Polymarket earnings bets that beat the odds on KPMG's clients: a Bitquery investigation that scored a group of wallets the same way.

Plans and the trial are on the Bitquery MCP server page.

Frequently Asked Questions

How do I track Polymarket whales?

Rank wallets by traded volume with polymarket_top_traders, remove market makers by looking at order counts and the number of markets in polymarket_trader_summary, then read each remaining wallet's bets and profit with polymarket_trader_activity and polymarket_trader_positions. The Bitquery MCP server runs these from a chat in Claude, ChatGPT or Cursor.

Are the top Polymarket traders by volume the most profitable?

No. Most wallets at the top of a volume ranking are market makers. In the example on this page, the eighth-largest trader of the week won three of six bets and was down about $434,000 before fees.

Can I copy trade with this?

Not directly. The server is read-only and places no orders. It covers the research step: which wallet is a bettor, what it holds and whether it has been winning.

How is this different from the Polymarket leaderboard?

A leaderboard is a fixed list. Here you ask follow-up questions about any wallet or market: lifetime history, positions with profit and loss, single fills above a size, and the wallets active in one market.

Can it alert me when a whale trades?

A chat does not run in the background. For alerts, use the Bitquery Polymarket GraphQL API with a WebSocket subscription, which carries the same data.

Do I need an API key?

No. You add the server URL to your client and sign in once with a Bitquery account in the browser.