The most reliable fingerprint of one operator running many wallets.
Every trade pays a base gas fee plus a priority fee the trader sets. Bot operators run all their wallets through one terminal with one fee configuration, so their wallets pay the exact same fee. Luminos groups the top holders by that shared fee fingerprint and shows how much supply each group controls. The fee shown for each group is the priority fee — the same number trading apps like pump.fun display — so you can match it against what you see in your own app.
Which trading terminal each wallet uses.
Terminals (Axiom, Trojan, gmgn, Photon, BullX) each route their fee to a distinct on-chain address, so we can tell which one a wallet trades through and tag it with a colored badge. The Accounts section sums this up: what share of the analyzed wallets — and of supply — sits on each terminal.
One result, and two numbers that mean very different things.
The header shows the result (Bundled / Cabaled / Organic / Unverified), a big number, and a confidence meter. A Bundled result also carries a small Fully or Soft tag — how deep the bundling goes (Fully = a hard trigger or the lowest CEX band; Soft = a milder band). Only Bundled ever shows a tag.
There are two separate 0–100 numbers, and mixing them up is the most common mistake:
| Number | Where | What it means |
|---|---|---|
| The big score | verdict header | Category-fit strength — how strongly the coin fits its result (0 weak → 100 strong). Not risk, not safety, not a price call. Read it with the color. |
| Risk score | Engine breakdown | The actual rug-danger number. Fed almost entirely by the bundling axis (~80%) plus a few scam flags. Cabaled and Organic add nothing to it. |
Unverified means the engine declined to grade the coin, for one of two reasons: it has no real market yet (almost no liquidity, few holders, barely any trades), or exchange-funding data was unavailable so bundling couldn't be checked. One important rule: a coin that would have been Organic is downgraded to grey Unverified whenever the exchange-funding data is unreliable — the tool refuses to hand out a clean blue result it can't stand behind. The banner says it plainly: not a clean bill of health. Missing data only ever pushes toward caution, never away from it — a Bundled or Cabaled result earned from other signals is never softened by it.
How much data backed the result.
High / medium / low reflects how many independent data sources returned real data for this token (there are three: market data, holder/exchange data, and trading-terminal data). A brand-new token with thin history gets low confidence even when the signals look clear — treat those reads as provisional.
How much of the tradeable supply traces back to real exchange withdrawals.
This is the one people ask about most. Luminos traces the top holders' wallets back to see how much of the free float (the supply that can actually trade — see the glossary) was funded by withdrawals from real centralized exchanges like Coinbase and Binance. Real exchange money is the footprint of many independent retail buyers. The ideal is a band, not a maximum — too little looks manufactured, and suspiciously high looks like one operator routing everything through exchanges to look organic.
| CEX funding (of free float) | Reads as |
|---|---|
| under 40% | Bundled — too little real exchange money (under 30% is "fully" bundled) |
| 40–50% | Cabal-leaning |
| 50–75% | Organic — the healthy / ideal band |
| 75–85% | Unusually uniform funding — caution |
| over 85% | Bundled — funding is too uniform to be real retail |
So the short answer: 50–75% is ideal. Above 75% the coin can still stay Organic only if a withdrawal-batch check confirms those exchange withdrawals weren't siblings from one coordinated batch.
Cluster supply vs Partition. The funding panel has a toggle. Cluster supply can add up to more than 100% because one wallet can trace to several exchanges and is counted under each — it's the fuller picture of a venue's footprint. Partition assigns every wallet to one exchange and counts it once, so it's the clean, no-double-count split. Use "open bubble map" to inspect the full graph.
Not every exchange is judged the same. Binance and Coinbase are where nearly everyone routes, so their normal range is wide (roughly 10–35% is fine) — and a very low share is itself a small flag (the coin avoided the mainstream venues). Regional or low-trust exchanges, and fiat/privacy on-ramps, get flagged at much lower shares because heavy funding through them is abnormal. Moonpay (a card on-ramp) is the special case that can flip a coin by itself: over 10% is at least Cabaled, over 20% is Bundled.
Supply first funded through a privacy mixer or instant-swap.
Right next to CEX funding, this shows how much of the top-holder money arrived through a mixer or instant-swap service (ChangeNow, FixedFloat, THORChain…). Legitimate exchange withdrawals are traceable; mixer money is deliberately not — it's used to hide where the supply came from. A separate line calls out when the deployer's own wallet was funded that way.
Groups of wallets that behave as one entity.
Wallets that bought in the same blocks, share a common funder, or hold near-identical amounts are grouped into clusters. Time nodes are the related idea in time: wallets that all first became active in the same tight window right after launch (co-buying). Coordinated buyers move together in the first seconds; organic buyers arrive spread out. Three separate detectors look for this — connected wallet clusters, timed entry bursts, and shared hidden-fee groups — and they share one scale:
| Supply held by one coordinated group | Reads as |
|---|---|
| under 5% | Normal |
| 5–20% | Insider coordination (cabaled) |
| over 20% | Manufactured distribution (bundled) |
Where the supply actually sits, and how the wallets are shaped.
Tight, uniform, freshly-created holdings are the shape of a manufactured launch; a wide spread of ages and sizes is the shape of an organic one. The at-a-glance cards each carry their own healthy-to-flag bands:
| Card | Healthy | Caution | Flag |
|---|---|---|---|
| Top-10 concentration | <30% | 30–50% | >50% |
| Fresh wallets (made near launch) | ≤20% | 20–30% | >30% |
| Age concentration | <25% | 25–50% | ≥50% |
| Identical holdings | <10% | 10–30% | ≥30% |
| Zero-sell top 10 (never sold) | 0–1 / 10 | 2–4 / 10 | ≥5 / 10 |
Age concentration and identical holdings are clustering scores (higher = more bunched-up), not a plain "X% of wallets are identical." All of these are measured against free float.
Healthy vs concerning, at a glance.
| Signal | Healthy | Concerning |
|---|---|---|
| CEX funding | 50–75% | <40% or >85% |
| Mixer funding | <3% | >10% |
| Top-10 concentration | <30% | >50% |
| Fresh wallets | ≤20% | >30% |
| Coordinated cluster | <5% | >20% |
| Largest time-node cohort | <5% | >20% |
| Shared gas-fee group | <5% | >35% |
50–75% of the free float. That band is where funding looks like many independent retail buyers coming off major exchanges. Below 40% reads Bundled (too little real exchange money); 40–50% is cabal-leaning; and — counter-intuitively — above 75% is a caution and above 85% reads Bundled again, because near-total exchange funding looks like one operator staging it to seem organic. More is not better.
It's the middle result: insider or exchange-routing coordination without the harder signs of bundling. Think one exchange dominating the funding, money skewed through regional / low-trust / fiat on-ramps, or a small number of wallets holding most of the float. It's a genuine caution — a signal that a coordinated group can move the coin — just a different, usually milder, pattern than Bundled. Note it doesn't feed the numeric risk score, so don't read "low risk number" as "safe" on a Cabaled coin.
Depends entirely on the color. The big number is how strongly the coin fits its result, not how risky it is. High on Organic = confidently clean; high on Bundled = a confident bundling call. The actual rug-risk number lives separately in the engine breakdown.
No. Organic means "no manipulation pattern found in the data we could check" — it is not a safety rating and not a price prediction. It's blue, not green, exactly so it doesn't read as a buy signal. A coin can be Organic and still lose most of its value, or Bundled and trade normally for a while. Always do your own research.
The engine couldn't grade it — either there's no real market yet (almost no liquidity or holders), or exchange-funding data was unavailable so bundling couldn't be checked. It's not a pass and not a fail. Re-scan in a minute; if the coin is brand new, wait for a bit of trading history first.
A result reflects the on-chain data at scan time, not a permanent label. New trading, new holders, or corrected upstream data can move it. That's expected — re-scanning is how you get the current read.
No. Every result is an automated, algorithmic opinion from public on-chain data — not a finding of fact, a fraud certification, or a claim about anyone's intent. If a project disagrees, there's an appeal link on every result and on the methodology page.