Every input, every weight, and every limit. If you disagree with the method, you can see exactly where.
Any tool can tell you what is being bought. Almost none tell you who is buying, and whether that who is real.
Four wallets buying the same token in the same minute can be four independent people who each did their own work. Or it can be one person with four addresses, manufacturing the appearance of demand. On a chart these look identical. The difference decides whether the token runs or whether you are the exit.
Finding that out by hand takes about twenty minutes of block explorer work per token, and by then the entry is gone.
Nothing here is bought from a data vendor. Every input comes from scans FINNPUTER ran itself.
Every scan FINNPUTER has ever run wrote down the market cap at that moment. A token scanned at $30k that trades at $900k today ran 30x, and we can prove we saw it early because we recorded the before-price ourselves.
For each of those tokens, the full swap history is walked and the earliest buyers are ranked by when they first bought.
Many early holders never paid for anything. They were handed the bag by the deployer. Tokens that arrive without payment leaving the wallet are not a trade, and those wallets are labelled and excluded.
Every wallet is cross referenced against our own cluster database. A wallet that is early everywhere and appears inside bundles is a sniper bot. Its numbers look excellent by design. It is capped and can never be recommended, whatever the arithmetic says.
A trade counts as won only when the wallet actually sold back. An open position showing 40x on paper earns zero points here. Paper gains are a screenshot, not a result.
Buying at $30k and buying at $3M are not the same skill. Average entry market cap is reconstructed from what the wallet actually paid per token, and it moves the score.
Wallets that clear the bar go on a watchlist. When several of them, with no connection to each other beyond our own rating, buy the same token inside the same window, that is the signal. Only then does the safety engine run.
One wallet can be lucky. A group that keeps showing up early together is a pattern.
Across every token we scanned, wallets that were among the first buyers of the same tokens, again and again, within the same window, are grouped. Counting shared purchases, not guessing at relationships.
Where several wallets in a group were funded from the same source, they are treated as one operator with several addresses. Their siblings, funded the same way, are checked too.
A group is judged on every token it was early in, not only the ones that ran. A group that buys everything lands near the market average and is worth nothing as a signal, however many winners it touched.
A group that clears the bar is frozen on a date. From then on it is measured only on tokens that did not exist when it was chosen. That is the only number we treat as proof.
Some groups are early in tokens that almost always die. They are frozen as warnings. When one of them is early in a token, the call is held back.
Every morning the latest runners are searched for new wallets and groups, with the same bar as by hand. Groups that are mostly copies of an existing one are not added twice.
A good record is only useful if you can follow it.
Almost always among the first five buyers and out within minutes. The numbers can look excellent. Nobody can follow it, because it has already left when any call goes out.
Early, and usually out inside an hour. Followable when the call is fast, which is why a call goes out within minutes of the buy.
Buys after the first rush and stays for an hour or longer. Still in when the call arrives. These are the wallets the signals lean on.
The profile comes from where in the line of buyers a wallet usually is, and how long it held, measured on the tokens where we saw both the buy and the sell. Too few measured holds means no verdict, not a guess.
Every call is ranked before it goes out.
Our strongest wallet groups or top wallets are early, or several rated wallets are buying it together right now, or one of our top rated wallets just bought.
Passed the safety engine and the signal bar, without that wallet evidence. Most calls are B.
A warning group is early and no good one is. These calls are not sent at all, and so they do not appear in any record.
The exact rules and which wallets sit in which group are not published. The results are: every tier is measured from the market cap at the moment of the call, misses included.
Puter Score, 0 to 100.
| COMPONENT | RANGE | WHAT IT MEASURES |
|---|---|---|
| Base by label | 0 to 60 | proven trader, repeatedly early, mixed, once, sniper, gifted |
| Realised profit | 0 to 20 | net returned, and average multiple on closed trades only |
| Entry quality | 0 to 10 | average market cap when they bought |
| Consistency | 0 to 10 | early across several tokens, not once |
| Rug exposure | 0 to -25 | share of touched tokens our scanner later flagged |
| Bundle association | 0 to -30 | appearances inside detected clusters |
Smart money 35, safety 35, liquidity 15, momentum 15.
Live mint authority, live freeze authority, or liquidity below the floor caps the total no matter how many good wallets are buying. Smart money being early in a rug happens constantly.
The part most tools leave out.
A wallet that made a fortune in tokens FINNPUTER never scanned is invisible here. Coverage grows with every scan. It will never be complete, and a score built on thin evidence is marked as such rather than hidden.
The score says a wallet was early and closed in profit. It does not say that repeats. Nobody knows that, including us.
When funds move to a centralised exchange or across a bridge, the chain of custody stops. No tool solves this. Most do not say so.
It is derived from what was paid per token against current supply. If supply was burned since, the figure drifts. Directionally right, not exact to the dollar.
Watched wallets are polled on a rotation, so a cluster can be a few minutes old before it appears.
A group found in earlier winners will look strong on those same winners. Only tokens that appeared after the group was frozen say anything about the future, and that sample is small at first.
Each call is hashed when written. That proves the record was not edited afterwards. It does not prove when it was written. On-chain anchoring is a later step.
Not because it is difficult. Because it needs history you cannot buy.
Contract checks, holder lists and liquidity data are available from vendors, which is why every tool has them and why none of them are a differentiator. Wallet level provenance needs your own record of what you saw, and when, going back months. FINNPUTER has been scanning and storing since early 2026. That archive is the product. Everything on this page is a way of reading it.
This site is free and needs no account.
In the bot, trading costs no credits. There is a trade fee and nothing else. Credits exist for one reason: when the AI does work for you, that work costs real money to run, and it would otherwise be free to drain from the outside.
Starter credits unlock once your wallet has been funded, which keeps throwaway accounts from farming them. Topping up is one tap from the wallet you already trade with.
FINNPUTER is an instrument. It reports what it measured and marks what it could not. A high score is not a recommendation, a low risk number is not a guarantee, and a token can fail for reasons no scan detects. Verify addresses yourself and risk only what you can lose.