Quick answer
To check whether KOLs are dumping a token, count how many tracked holders sold recently as a share of everyone who held it, then look at who sold and what they realized. On E/ACC, 7 of 105 tracked traders (6.67%) sold, under a 10% alarm line. The catch: the tool covered only about 2 hours of the 24 requested.
How do you tell if KOLs are dumping a token?
You measure the sellers against the whole group. One famous trader selling is an anecdote. Seven of 105 tracked traders selling inside the same window is a rate you can compare across tokens and across days.
StalkChain's stalkchain_fomo_kol_sell_pressure does that arithmetic. It takes a token address, looks at recent sell alerts from tracked traders, and returns the share of holders that sold, who they were, what they realized, and who still holds.
This guide pulls it live on E/ACC (Solana), reads every field, and spends as much time on the limits as on the signal. The limits are the part most "dump detector" pages skip.
What does the call look like?
In Claude or ChatGPT with the StalkChain MCP connected, the prompt is plain English:
Are tracked KOLs dumping E/ACC on Solana? Check the last 24 hours.
Under the hood the assistant makes this call:
stalkchain_fomo_kol_sell_pressure
address: CbcyNo7m1amFWqEQm2m4PLv1UNvpcL3C1Ujm6AkzpKoU
windowMinutes: 1440 (default, 24h)
thresholdPercent: 10 (default)
Pulled Oct 8, 2026, about 06:14 UTC via StalkChain's MCP (stalkchain_fomo_kol_sell_pressure). Trimmed response:
{
"data": {
"token": "CbcyNo7m1amFWqEQm2m4PLv1UNvpcL3C1Ujm6AkzpKoU",
"windowMinutes": 1440,
"coveredFrom": "2026-10-08T04:23:15.000Z",
"currentTrackedHolders": 100,
"sellersInWindow": 7,
"fullyExitedSellers": 5,
"trimmedButStillHolding": 2,
"unmatchedSellers": 0,
"denominator": 105,
"sellersPct": 6.67,
"thresholdPercent": 10,
"thresholdCrossed": false,
"sellers": [
{ "handle": "FreeLateDog", "lastSellAt": "2026-10-08T06:00:05.000Z", "realizedPnlUsd": -202826.48, "stillHolds": false },
{ "handle": "s226", "lastSellAt": "2026-10-08T05:28:25.000Z", "realizedPnlUsd": -40909.51, "stillHolds": false },
{ "handle": "MingusCapital", "lastSellAt": "2026-10-08T04:42:30.000Z", "realizedPnlUsd": -40783.7, "stillHolds": false },
{ "handle": "Dxranteth", "lastSellAt": "2026-10-08T05:36:25.000Z", "realizedPnlUsd": 37612.63, "stillHolds": true }
],
"stillHoldingTop": "[15 entries trimmed: smokey0x 57123.02, free2trade 35894.44, Dxranteth 35853.97 ... oznek 14125.55]"
},
"notes": [
"Sell feed hit the 100-alert cap; sells older than 2026-10-08T04:23:15.000Z are not visible."
]
}
Three of the seven sellers are shown here, plus Dxranteth because he is the interesting case. The full response lists all seven.
Same call over REST:
curl -G https://data.stalkchain.com/api/v1/tools/stalkchain_fomo_kol_sell_pressure \
-H "Authorization: Bearer sc_YOUR_API_KEY" \
-d address=CbcyNo7m1amFWqEQm2m4PLv1UNvpcL3C1Ujm6AkzpKoU \
-d windowMinutes=1440 \
-d thresholdPercent=10
How do you read the payload field by field?
windowMinutes and coveredFrom. We asked for 1,440 minutes. The tool could only see sells from 04:23 UTC onward. Read these two together before you read anything else.
currentTrackedHolders: 100. This is the number of tracked traders holding the token now, and it is a ceiling. The tool will not report more than 100 tracked holders for one token, so on a widely held token the count tops out.
sellersInWindow, fullyExitedSellers, trimmedButStillHolding. Seven sellers: five sold out completely, two sold part and kept the rest. These are different behaviors and should not be lumped together.
denominator: 105. A trader who sold everything is no longer a current holder, so the tool adds them back: 100 current holders plus 5 who fully exited. Skip that step and the percentage would look worse than it should.
sellersPct: 6.67 and thresholdCrossed. 7 divided by 105. The default alarm is 10%, and you can change it with thresholdPercent. Nothing is magic about 10%. It is a default, so pick your own and keep it consistent across tokens.
realizedPnlUsd. The profit or loss the alert attached to that sale. Positive and negative values mean very different things, covered below.
stillHolds. Whether the seller still has a position. This is how you separate an exit from a trim.
unmatchedSellers: 0. Sellers whose alert could not be matched to a current holder or an exit. Zero here, but watch it on other tokens.
What does the coverage window mean for the result?
It means 6.67% is a floor, not a 24-hour answer. The tool states its limit itself: coverage is bounded by the newest 100 sell alerts, and the response says which time range it actually covered. Here that was about two hours.
So the honest reading is: in the roughly two hours the tool could see, about 7% of the tracked group sold. Nothing in this response says what happened in the other 22 hours. A true 24-hour figure could be higher.
This is why a quiet result is weak evidence. "No dump detected" and "the tool only saw two hours" can produce the same output. A tool that tells you which one you got is doing you a favor. Make a habit of checking coveredFrom first.
If you need the number for a specific shorter window, ask for it. A 60-minute request on a fast token can be covered in full, and then the percentage is a real 60-minute percentage.
Is a negative realized PnL a sign of dumping?
Not by itself. FreeLateDog sold out at a realized loss of about $202,826. That reads as a trader cutting a losing position, which is capitulation, and it tells you about that trader's cost basis rather than about a coordinated exit.
Compare Dxranteth: a realized gain of about $37,613 on a partial sale, and he still holds roughly 11.6M tokens. That is a trim, profit taken while keeping exposure.
A rough guide to the two shapes:
- Sellers mostly realized losses and fully exited. Often stop-outs after a drop. Pressure may already be spent.
- Sellers mostly realized gains and kept part of the position. Profit-taking, which can continue as price climbs.
- Sellers realized large gains and fully exited. The pattern worth a closer look, especially when several appear close together.
None of this proves intent. A tracked trader's sale is one wallet's decision, and tracked traders are a small slice of all holders.
How do you confirm who still holds?
Pair the sell-pressure call with stalkchain_fomo_kol_holders, which lists who holds now. On E/ACC at the same pull it showed 100 tracked holders with about $649,498 in tracked value, and the top five holding about 30.19% of that tracked value. The largest, smokey0x, held about $57,123.
It also returns 24-hour flow across all wallets, not just tracked ones: 5,331 buys and 4,520 sells, with about $776,966 bought and $825,716 sold, a net of about -$48,750.
The one-hour buy/sell ratio of 1.99 is a count of trades, while the one-hour net volume was negative (-$8,108). More buyers by count, more dollars leaving. Do not read 1.99 as bullish without checking the dollar side.
One honest note: the holders payload also reports a total holder count, and another StalkChain tool reports a very different number for the same token (18,628 versus 7,523 at this pull). We skipped that field entirely rather than pick one. Where two sources disagree, show both or show neither.
What are the false positives to avoid?
Trim versus exit. stillHolds: true means a seller kept a position. Count it, but weigh it less.
A short window read as a verdict. See the coverage section. A 2-hour sample cannot support a 24-hour claim.
Tracked traders are not the market. Seven tracked sellers can sit beside thousands of ordinary sellers. The 24-hour flow above shows 1,851 unique sellers; the tracked group is a tiny part of it.
Losses are not all signals. A trader selling at a loss is often forced or fatigued, not informed.
Holding is not conviction. A tracked wallet can sell in the next block.
Does this replace a liquidity or safety check?
No. Sell pressure tells you what a group of traders is doing. It does not tell you whether you can exit. Before acting on any token, check the exit side with liquidity-removal and exit-risk signals, and compare against the opposite signal, whale accumulation.
If KOLs are buying while a few sell, that is a different story from KOLs and whales both leaving.
For the identity side of the tracked group, see how Solana KOL tracking resolves handles to wallets, and the wider smart money wallet guide.
How does this compare with other tools?
GMGN has a smart-money exit signal skill page, which is a reasonable place to see a quick directional read. When we reviewed it, it did not publish a sample payload or describe its sample window. See the GMGN smart money exit signal page and judge for yourself.
The difference we lean on is narrow: this tool returns realized PnL per seller, states its coverage window in the response, and works inside an AI assistant over MCP. It is not a promise of better signals.
How do you run this yourself?
Connect the StalkChain MCP server to Claude or ChatGPT (setup for Claude is on the site), then ask the plain-English question above. Everything here is also covered on the Solana data page. Plans and the free starter credit are on the pricing page.
A simple routine: run it with the default 24-hour window, read coveredFrom, rerun with a shorter window if coverage was truncated, then check holders and exit liquidity before drawing any conclusion.
Data, not financial advice. This article shows how to read a tool's output on one token at one moment.
FAQ
Are KOLs dumping this token right now?
Check the share of tracked holders that sold in the covered window, then who sold and whether they kept a position. On E/ACC at 06:14 UTC on Oct 8, 2026, 7 of 105 tracked traders (6.67%) sold, below the 10% default threshold, over roughly the last two hours.
Why did it only cover two hours when I asked for 24?
The tool reads the newest 100 sell alerts and filters them to your token, so a busy feed shortens the window. The response states the real range in coveredFrom and a note. Treat the percentage as a floor for the longer period.
What does the 10% threshold mean?
It is the default thresholdPercent at which the tool flags thresholdCrossed. It is a convention rather than a market rule. Set your own and apply it the same way to every token you compare.
Does a seller with a loss mean a dump?
Not on its own. A seller who exits at a loss is often cutting a bad position. Large realized gains with a full exit, clustered in time, are the pattern to look at harder.
How many tracked holders can it see?
Up to 100 per token. On widely held tokens the tracked count tops out, so it describes the tracked group, not all holders.
Can I get alerted when tracked traders buy a token?
Yes, for buying: StalkChain alerts can fire when 2 or more tracked traders buy a token. They check about every 5 minutes. See which tokens several traders just bought.