research updated 2026-06-30 building

NFTX protocol — overview & ideas

Research note · 2026-06-30 · Source: NFTX v4 whitepaper (a Flayer Labs product, part of the FLAY ecosystem)

Read for what’s worth borrowing into Agents4.fun. Section 5 marks the ideas — they’re prompts, not commitments.

What it is

NFTX is a “fungibility layer” for NFTs: deposit an NFT into a collection’s pool and you get back a fungible ERC-20 floor token (e.g. MILADY) that trades on an AMM, so an otherwise illiquid JPEG becomes something you can buy, sell, or swap instantly. The classic version (2020, on Uniswap V2/V3) only made the floor liquid. The v4 relaunch — the subject of this whitepaper — rebuilds the protocol on Uniswap V4 and extends fungible liquidity to every rarity in a collection, not just the floor, via a self-assessed listing model. The pool contracts are non-upgradeable and feeless to enter/exit; the protocol earns through listings, not deposits.

How it works

Vault → floor token. Each collection has one Uniswap V4 pool pairing the collection’s floor token against ETH (specifically flETH, an ETH wrapper that earns staking yield — see below). Depositing any item from the collection mints 1 floor token, which you can immediately sell on the AMM for floor value. Redeeming a floor NFT from the pool burns 1 token. Entry and exit are free.

Liquid Listings (the key v4 idea). Instead of every deposit collapsing to the floor, a seller can deposit a non-floor item and self-assess its price as a multiple of floor — e.g. list a rare Milady at 1.3x. They immediately receive 1 floor token (instant ~75% of value) and keep a claim on the remaining 0.3 that pays out when the listing is filled. So value splits into “instant liquidity now” + “listed upside later.”

Harberger-style listing fees keep prices honest. Listing isn’t free — you pre-pay a fee in floor tokens, deducted from the liquidity released at listing. The fee scales with the square of the floor multiple and linearly with duration (normalized to a 7-day reference). Doubling your ask costs ~4x more to maintain; a kink at 2.00x softens the curve so “grail” listings stay viable. Because anyone can fill or re-list at your self-assessed price, there’s strong pressure to price fairly — overpricing just bleeds fees until the listing closes. Unused fee is refunded on early close.

Dutch auctions on expiry. If a listing’s pre-paid balance runs out without being filled or closed, the item auto-rolls into a Dutch auction that decays linearly from its current multiple down to floor (1x), where it becomes a plain redeemable floor NFT. No keeper needed — the expiry is known up front. Sellers can also send an item straight to a Dutch auction (a “Dutch Listing”) for fast liquidation — useful for NFT-Fi protocols liquidating collateral.

Re-Listing (appraiser arbitrage). Anyone can re-list an item already in the pool that they don’t own. If it sits at floor, you re-list it higher by paying only the new listing fee — no need to buy it first. If it already carries a listing, you pay the current premium to its owner plus the new fee. This lets people with trait knowledge profit by repricing mispriced assets without inventory risk.

Trade-Ups. Because listings are priced in floor-token multiples, a rare listed at 3x can be filled with 3 floor items (or 2 floors + ETH). This natively enables floor→rare swaps and, across collections, cross-collection trade-ups/down.

LP rewards via donate(), not staking. This is the biggest mechanical departure from old NFTX. There is no LP-token staking. Protocol revenue (listing fees + swap fees) is pushed directly into the V4 pool via Uniswap’s donate() method, inheriting the same fee-claim logic as swap fees. LPs collect their share passively, no extra gas or staking step. These “exogenous fees” are real yield on top of swap fees, partially offsetting impermanent loss and letting LPs quote tighter spreads. A configurable protocol-fee share can be carved out first; set to zero, 100% of fees go to LPs. Users/creators can also bribe LPs via donate() to bootstrap a collection’s liquidity.

Other pieces. flETH is an ETH-equivalent wrapper whose underlying ETH earns staking yield routed to a protocol receiver (NFTX LST). Collections enter through a swappable launch gate (permissionless or curated) and can launch single-sided (token-only, lower capital). A permissionless shutdown path lets the last holders of a dying collection lock-and-vote to liquidate remaining NFTs and split ETH pro rata. An onchain price oracle from pool observations feeds NFT-Fi, and a zap bundles “buy NFT with ETH” into one tx.

The good

Instant liquidity for illiquid assets. The core win: turn a JPEG into a fungible token you can exit immediately, no waiting for a bidder.
Liquidity across every rarity, not just floor. The self-assessed Liquid Listing model is the genuine innovation — mid-tier and rare items get a market, not just the floor.
Honest pricing by design. Harberger-style self-assessment + squared fee curve + permissionless re-listing creates real economic pressure to price fairly, without an oracle or appraiser committee.
Passive, real yield for LPs. Routing protocol fees through donate() means LPs earn beyond swap fees with zero staking ceremony — cleaner UX and a partial IL hedge.
Composability. ERC-20 floor tokens + an onchain oracle + zaps make NFTX a Lego brick for NFT-Fi (lending, collateral, structured products).
Credible neutrality. Non-upgradeable pool contracts, feeless entry/exit, limited admin surface — fewer rug vectors, more integratable.
Clean exits even for dead collections. The permissionless shutdown path means the last holder isn’t trapped.

The bad / risks

Fungibilization still flattens trait value at the floor. A floor token represents a member of the collection, not your specific one. The Liquid Listing layer mitigates this but only while a listing is live and fairly priced; the underlying pool token is still floor-grade fungibility.
Random / non-targeted redemption at the base layer. Redeeming a plain floor NFT gets you a pool item, not a chosen one. Targeted value only exists through the listings layer, which costs fees and time.
Peg / depeg dynamics. The floor token’s value depends on the pool actually holding fairly-priced, redeemable items. Extortionate listings, thin liquidity, or fee mispricing can decouple token price from real floor — the fee model fights this but doesn’t eliminate it.
Fee complexity. Squared, duration-scaled, kinked listing fees are powerful but opaque to ordinary users; mispricing a listing can quietly bleed value into fees and dump the item into a Dutch auction.
Governance / centralization residue. “Non-upgradeable pool” still pairs with a protocol owner who registers managers, sets gates, configures fees, and holds a safety pause. Neutral-ish, not trustless.
Market & regulatory risk. Fungible NFT tokens, staking-yield ETH wrappers, and pooled redemption sit in an unsettled regulatory zone. And the NFT market itself has cooled hard since the last cycle — a liquidity layer is only as useful as the demand beneath it.
Smart-contract / dependency risk. Deep coupling to Uniswap V4 hooks, flETH staking strategy, and external liquidation venues — more surface, more trust assumptions.

Innovation worth borrowing — for Agents4.fun (ideas, not commitments)

Agents4.fun is a whitelist + sweepstakes product on Ethereum with NFTSweeps mechanics folded in, and it draws on DePunks’ NFT mechanics + on-chain PRNG. NFTX isn’t a sweepstakes protocol, but several of its primitives map cleanly onto that surface. Marked as ideas:

Self-assessed pricing as an entry mechanism. NFTX’s Harberger listing — “name your price, pre-pay a fee that scales with how aggressive you are” — is a clean anti-spam / fair-entry primitive. Idea: let sweepstakes entrants self-assess a stake or bid, with a squared fee curve that prices out trolls while keeping genuine high-rollers viable. The kink-above-2x trick (soften the curve for whales) is a nice knob for VIP whitelist tiers.
Vault → fungible ticket token. NFTX turns one NFT into one fungible token. Idea: a sweepstakes “pool” where deposited NFTs (or ETH) mint fungible entry tickets — tradeable, stakeable, composable — so people can buy/sell their odds before the draw, and secondary liquidity for entries becomes its own market. This is directly the NFTSweeps-into-Agents4.fun fold.
Random redemption RNG ↔ sweepstakes draw. NFTX’s “redeem a random floor item from the pool” is mechanically the same problem as “draw a random winner.” DePunks already has on-chain PRNG; reuse it as the shared randomness engine for both the random-pack/redemption flavor and the sweepstakes draw. One audited PRNG, two products.
Staking-for-fees instead of staking-for-emissions. NFTX’s donate() model — real protocol revenue flowing to participants passively, no LP-token lockup, no inflationary emissions — is the sustainable yield pattern. Idea: whitelist holders or ticket stakers earn a share of real sweepstakes fees/rake routed automatically, not a printed governance token. Stronger story, fewer regulatory and tokenomics headaches.
AMM-for-NFTs ↔ AMM-for-odds / whitelist spots. If whitelist spots or sweepstakes entries are fungible tokens, an AMM gives them instant price discovery — a live, market-clearing price for “how much is a whitelist spot worth right now.” Idea: a thin AMM (or even a Dutch-auction allocation) for whitelist allocation, so spots aren’t first-come-gas-war but priced.
Dutch-auction allocation for whitelist / mint. NFTX’s keeperless Dutch auction (decay from a multiple down to floor, expiry known up front) is a tidy, gas-efficient allocation mechanism. Idea: allocate scarce whitelist slots or sweepstakes entries via a Dutch auction — fairer than gas wars, no keeper bots, and it self-prices demand.
Permissionless re-listing ↔ agentic appraisal. NFTX lets anyone re-price a mispriced pool item for profit, no inventory risk. Idea: this is a natural agent job — an Agents4.fun bot that monitors entries/listings, spots mispriced odds or undervalued whitelist spots, and re-prices them, earning the spread. Ties the “agentic” thesis directly to a concrete revenue loop.
Permissionless clean shutdown. NFTX’s lock-vote-liquidate-distribute wind-down is a good model for ending a sweepstakes round trustlessly: lock entries, reach quorum/close, draw, distribute pro rata, sunset the round. Borrow the pattern for round lifecycle, not just collection death.