Twenty wallets.
How many operators?

01The score

A number from 0 to 100, where 100 is the cleanest distribution. It adds up five parts, each measured on the top 20 holders against the real float: supply outside the curve, pools, lockers and burn addresses.

Between the two marks a part scales in a straight line. Bands: CLEAN 80 and up, OK 60–79, RISKY 40–59, DANGER under 40.

02Red flags and caps

Some patterns are too clear for a weighted sum. These rules override it.

Red flags: the verdict is DANGER whatever the score

• An operator of two or more linked wallets could move the price down 50% or more by selling.
• 80% or more of the read top wallets never traded anything before this token.
• 10% or more of the float was received by transfer instead of bought.

Caps: the verdict is at best RISKY (score 59 or lower)

• A proven operator of 3 or more wallets, or a pack of fresh wallets, holds part of the float.
• One independent holder with a trading history could move the price 50% or more. One honest whale is a risk, not a rug.
• 3 or more of the top 20 wallets could not be read in time.
• Liquidity is under $1,000.

The sell scenario on the chart only appears on DANGER verdicts caused by linked, fresh or transferred supply, and only when selling it would drop the price 40% or more. It never changes the score.

03When there is no score

• TOO EARLY Fewer than 10 wallets hold the token outside the curve. With so few holders one buy changes everything, so the scanner waits.
• TOO ESTABLISHED Older than 30 days with high liquidity and market cap. On such tokens the top holders are mostly exchanges and big pools, and holder patterns stop meaning what they mean on a fresh launch. Owner powers and liquidity still apply.
• PARTIAL Some wallets could not be read within the time budget. They are marked unread and never counted as evidence; with 3 or more unread the verdict is capped at RISKY.

04Linked wallets

Proven links merge wallets into one operator: the same buy transaction, tokens received from the same ordinary wallet, and direct transfers between holders.

Packs are 3 or more fresh wallets that enter within a couple of slots with amounts within ±30% of each other. A pack is behaviour, not proof, so its share counts at 60%.

Exchanges, routers, bridges and any address with 100 or more outgoing transfers are never used as a link, so strangers who bought through the same app are not glued together.

05First buyers

The scanner reads the first 20 buyers after launch and records how many seconds after the launch each one bought and how much. Then it follows what each did since: still holding, added more, sold part, sold everything, moved the tokens, locked or burned them.

• Snipers: wallets that bought in the first 30 seconds. The share of float they still hold is the Unsold snipers part of the score, up to 10 points.
• Moved to one wallet: when several early buyers send their tokens to the same destination, they are grouped, because one person usually bought through many wallets.
• Locked or burned tokens are not counted as an exit.

On a token page the first buyers have their own table. Through the bot, a watched token sends an alert when its early buyers start to exit.

06Owner powers

Read from the mint account in one request, so the card appears before the holder scan finishes.

• Mint authority: if set, someone can create new tokens and dilute every holder.
• Freeze authority: if set, someone can freeze your token account so you cannot sell.
• Metadata: if mutable, the name, ticker and image can change.
• Transfer fee: on Token-2022 tokens, the fee on every transfer and whether someone can raise it.

07Liquidity

On the pump.fun bonding curve there are no LP tokens and nobody can withdraw the liquidity. PumpSwap burns LP tokens at graduation. On Raydium CPMM pools the right to withdraw sits in LP tokens, so the scanner reads who holds them: burned, locked or in a wallet. Other pool types are shown as not checked rather than guessed.

08Guard

Guard runs on every token page and in every scan. It reads the chain: launch-slot buys, dev holdings now and the creator's history. A launch is held back only on evidence:

• Creator history: the dev wallet launched 10 or more tokens and fewer than 1 in 5 are still alive;
• First-slot buyers: 30% or more of supply was bought in the launch slot;
• Dev holdings: the dev still holds 15% or more of supply.

On the Radar, the grid and the home page every new launch gets Shield, a lighter check made from free data: the creator's pump.fun record from the chain, read once the launch reaches 10% of its curve or 20 buyers (held at 10 or more launches with under 10% graduated), and the dev's buy at launch (held at 15% or more). A held launch is still shown on the Radar with its reason when you switch the Shield filter to show all. On a token page the Who made it card shows the dev wallet, how many tokens it launched, how many are alive, its best market cap, and each Guard check with its number.

09Limits

• The scanner only reads public on-chain data. It never asks for keys or seed phrases and never touches funds.
• It sees wallets, not people. Two wallets of one person with no on-chain connection look independent.
• A CLEAN score is not a promise. Liquidity, sentiment and the dev's off-chain plans are outside what it can read.
• Nothing here is financial advice.