The desk that reads coins gives its whole reading away. Paste a mint into the address below and one request returns the market, the contract authority, the holders, the linked-wallet clusters behind them, the genesis of the coin, the supply locks and the wallet they pay out to, and every burn with the transaction that proves it - each field carrying the source that produced it.
No key. No signup. No trial window. Nothing here is held back for a paid tier: the free answer is the whole answer.
The first URL is the rating flow itself: one request, and the agent's full written verdict comes back as JSON with engine in the response - model when a language model wrote it from the live blocks, desk when the composed fallback did. A generated answer and a composed one can never be mistaken for each other here.
/api/chat?mint=GxWqJbWPxMseev7WXYNPLEG8K59v2fAhJ5m5o7Yepumpone GET, the written answer, engine named in the JSONThe pack, for any mint/api/pack?mint=GxWqJbWPxMseev7WXYNPLEG8K59v2fAhJ5m5o7Yepumpevery block the answer is built from, with its sourceThe same read composed, no model call/api/chat?mint=GxWqJbWPxMseev7WXYNPLEG8K59v2fAhJ5m5o7Yepump&fast=1engine=desk, about a secondThe desk's own read of the same mint/api/agent/read?mint=GxWqJbWPxMseev7WXYNPLEG8K59v2fAhJ5m5o7Yepumpthe same numbers the agent answers fromHolder clustering, deeper/api/agent/insiders?mint=GxWqJbWPxMseev7WXYNPLEG8K59v2fAhJ5m5o7Yepumpevery wallet traced, linked or unresolvedEvery claim and the request that checks it/proofthe whole surface, one row per sentenceThe keyless index/apimachine-readable list of every doorFor PillCrew itself the pack carries 26,519,195 tokens burned across 8 transactions, each one openable on Solscan, and 3 lock contracts with their unlock dates and the wallet they return to.
One keyless GET returns the agent's full written answer for a mint, built from every block below. The response names its engine: `model` when a language model wrote it from the live data, `desk` when the composed fallback did. Nothing about that answer is canned - and nothing hides which produced it.
Price, market cap, liquidity, 24h volume and the change, with the vendor that measured it and the second it was captured. A snapshot older than the refresh window says so instead of pretending to be live.
Mint authority and freeze authority read from the mint account itself - the two switches that decide whether a coin can be inflated or frozen under you.
The enumeration, the pool vaults labelled from the live route, and the concentration with and WITHOUT those pools. A 50% treasury is never printed as one whale.
Each top wallet's earliest reachable transaction is read for a single-source SOL funding event. Wallets that share a funder, or funded each other, are grouped and summed; every wallet that could not be traced is published with the reason instead of being called clean.
The mint's own oldest transaction, decoded: who created it, when, what the authorities were in that same transaction, and the creator's opening buy as a share of the initial supply. Resolves for any launch whose history is short enough to walk - a coin minted years ago reports the gap instead of guessing.
Each Streamflow contract is followed to its escrow token account, the balance is read, the unlock dates are decoded from the contract bytes and so is the wallet the tokens return to. A lock is published as a schedule, never as a burn. Shown for a coin with a published lock set ($PILLCREW today); null elsewhere by design.
The burned total, the number of transactions and a Solscan link for every one of them. Permanent supply reduction, kept apart from the locks in the same answer. Shown for a coin with a burn ledger ($PILLCREW today); null elsewhere by design.
The third-party risk score and its flags, attributed to the vendor that produced them - so the desk's own read can be compared against an outside one in the same object.
What could not be measured this time, and why. A field we did not read is null with a sentence, never a zero that reads as 'none'.
Because a rating desk that hides its evidence is just an opinion with a chart. Every number this site prints is checkable, so the check is public: the pack, the clustering, the lock decoding and the burn ledger are the same code the agent answers from. If the answer and the pack ever disagree, the pack is right and the answer is a bug.
Nothing here is financial advice, and no field in the pack is a recommendation. It is a reading of the chain, published so you can disagree with it using the same numbers we used.