Docs

One transaction deploys the token, creates the pool, seeds it and locks the position. It is atomic, so a failure leaves nothing half-built: there is no state where a token exists without the liquidity behind it.

The entire supply of 1,000,000,000 goes into the pool single-sided, meaning the pad supplies only the new token and none of the asset it is priced against. That is what lets a launch open with no capital. There is nothing to match, and buyers arrive on the other side.

The resulting position is an NFT, and it goes to a vault with no function to withdraw it. Not a rule an admin could bypass, and not a timelock that expires: the vault never calls decreaseLiquidity and does not even declare it, so the principal has no path out. Fees are separate. They accrue to the position and can be collected without touching what backs the pool. Anyone may trigger a collection, because the money can only ever reach the two addresses fixed when the vault was made, split 50/50 between the creator and the treasury.

Every chain runs the same contracts against a different venue. All of them are Uniswap v3 or a fork of it, which is what makes fees claimable apart from the principal. A venue that folds fees back into the reserves cannot support this design at all, which is why some obvious candidates are missing.

Mantle is the exception worth understanding. Tokens there are priced against wrapped stocks such as wNVDAx or wTSLAx, not against MNT. That creates a problem, because nobody holds tokenised NVIDIA. Asking a buyer to go and get some before they can trade would be the end of it.

So a buy crosses three pools in one transaction, and the stock never reaches your wallet:

MNT  ->  USDT0  ->  USDC  ->  wNVDAx  ->  TOKEN
   Agni      Agni      Fluxion   Fluxion
   0.25%     0.01%      0.30%     1.00%

The first hops run on Agni and the last two on Fluxion. That split is forced. A Uniswap v3 path resolves every pool through one factory, so a single router cannot reach both venues, and the stocks exist only on Fluxion while MNT is deep only on Agni.

The gap is the whole reason. Fluxion’s MNT/USDC pool holds about $654, against $2.8M for Agni’s MNT/USDT0. Routing entirely on Fluxion cost about 13% in slippage on a $43 buy. Splitting it returns roughly 17% more tokens for the same MNT.

A contract called BuyRouter joins the halves. It takes your MNT, runs the Agni leg, then spends the result on the Fluxion leg and sends the tokens to you. It has no owner, no storage and no withdrawal function, it holds the intermediate USDC only between two calls inside one transaction, and the slippage bound is checked on the final output, which is the only number you care about. Selling reverses the route and unwraps at the end, so you get MNT back rather than USDC.

HyperEVM runs 3M-gas blocks by default. Larger blocks exist but are opted into per address through HyperCore, which a browser wallet cannot do. Creating a Uniswap v3 pool costs about 4.55M gas on its own, so a launch there could never build its own pool.

Pools are therefore created in advance, against token addresses that do not exist yet. The launchpad deploys tokens with CREATE2, so an address is known before the token is deployed, and a pool can be created for an address with no code at all. A launch then claims one and costs about 2.2M gas, which fits an ordinary block. If someone reserves one of those pools at a wrong price, the launchpad refuses it rather than seeding into it.

Every pool charges 1% per side. It accrues to the locked position and splits evenly between whoever launched the token and the treasury. The creator’s half can be redirected by the creator alone. The treasury’s half is fixed when the vault is created and cannot be pointed anywhere else, by anyone.