Uniswap
KLEAR3 ResearchPublished Updated
ArbitrumPolygon
X LayerUNIUniswap, the automated market maker that lets anyone swap tokens against a shared pool instead of a matching counterparty, has published code since April 2020. Since December 2025, a governance vote has routed part of its swap fees into buying back and burning UNI, and its own supply counter has not moved since.
What does Uniswap actually do?
Uniswap lets anyone swap one token for another against a shared pool of capital, not a matching counterparty. Liquidity providers deposit both assets and earn a cut of every swap; prices move automatically as each trade shifts the pool’s balance, with no order book and no market maker standing on the other side.
The organisation behind it describes itself, on its own GitHub page, as building “Web3 infrastructure and products” across 175 public repositories — a wider claim than “a decentralised exchange”, backed by a governance process, a token, and its own layer-two network rather than a single trading app.
Source GitHub, read 2026-10-15
That breadth is worth naming up front, because this page does not cover all of it. It covers the contracts in Uniswap/v3-core: the pool logic every swap on Ethereum mainnet still runs through, who has reviewed it, and what its token actually pays for.
Is the code actually open?
Published, yes; open source, not quite. GitHub detects no recognised licence on the code, so anyone can read it and nobody has been granted the right to reuse it. Two firms, ABDK and Trail of Bits, have reviewed the core pool contract; neither PDF report states exactly what they found.
The repository's licence field returns NOASSERTION, which is not a permissive licence and not a missing file: it is GitHub saying its detector could not identify one. A directory that prints a green licence badge here is filling in a blank, not reading one.
Source GitHub, read 2026-10-15
It is not the only core AMM to carry that exact gap: our profile of Pendle reads the identical blank, months apart, on its own core repository, read by us the same way.
ABDK and Trail of Bits each published a report here. Trail of Bits also left a README describing its method: three properties formally verified, all confirmed, on top of end-to-end fuzz testing of swaps, mints and burns.
Source Trail of Bits, read 2026-10-15
What neither PDF tells a reader, because we could not open either file this session, is how many findings each report actually raised, and how many of those were fixed before launch.
Source ABDK + Trail of Bits, read 2026-10-15
Is anyone still working on it?
Barely, by commit count, and that is not alarming for contract code meant to stay fixed. One commit landed in the 90 days to 15 October 2026, and it touches continuous-integration configuration, not the protocol itself: the core pool logic has not changed in that window at all.
That single commit's message reads “ci: restore scoped github_token so lint check can be created” — a repair to the repository's own testing pipeline, from one Git identity, not a change to the protocol itself.
Source GitHub, read 2026-10-15
A frozen core is not the same as an abandoned project. The organisation that publishes it keeps 175 public repositories and five publicly listed members, including its named founder — closer to a maintained company than an anonymous collective.
Source GitHub, read 2026-10-15
What shows it is running
- The UNI token contract holds deployed bytecode — 0x1f98…f984, read with eth_getCode, at block 26195627
- The core repository is barely moving, and the one commit is CI, not protocol — 1 commit in the 90 days to 2026-10-16
- The core repository made public
- Start of the 90-day window we measured
- Most recent commit at the time of reading
- This page, checked and published
Where does it actually run?
Everywhere, by one count: DefiLlama lists Uniswap deployments on 46 networks. Deposits are not spread evenly across them. Ethereum alone holds roughly two and a half billion dollars, more than the next four chains combined, including Unichain, the layer-2 its own publisher built and operates.
That concentration matters for the token discussed below, because the Ethereum contract is the one network we read ourselves: we asked a public node directly whether the UNI address holds deployed bytecode, and it does, at block 26 195 627.
Source Ethereum, read 2026-10-15
The network count and the ranking below it come from an aggregator we did not audit; this page says so rather than passing its tally off as our own measurement.
Source DefiLlama, read 2026-10-15
ArbitrumPolygonX Layer
MonadOP MainnetEach chip above matches a network DefiLlama lists for Uniswap deployments; the Ethereum one carries the extra weight of bytecode we read ourselves. DefiLlama also lists newer networks, including the publisher’s own Unichain, for which no official mark was retrieved this session.
What the token actually does
UNI pays for governance votes, and since December 2025 also captures a cut of swap fees for buybacks and burns. The contract’s own total supply still reads exactly 1,000,000,000, unchanged since 2020: burning moves tokens to an unreachable address rather than reducing that counter, which has not moved since genesis.
At genesis, in 2020, 60% of the one-billion-UNI supply went to the community, 21.266% to the team and 18.044% to investors, both vesting over 4 years; that vesting finished in 2024-09, and a perpetual inflation was designed to begin after that, at 2% a year, through minting governance can authorise.
Source Uniswap blog, read 2026-10-15
We read the contract ourselves rather than trust that design: its totalSupply() still returns exactly 1,000,000,000, the genesis figure, to the unit. The separate address where burned UNI actually lands holds 112,895,581.213 UNI, moved there, never subtracted from the contract’s own count.
Source Ethereum, read 2026-10-15
CoinGecko reads that dead balance as a reduction and reports total supply at roughly 887 million, with 625 million of it circulating. Neither figure is wrong; they answer different questions about the same token.
Source CoinGecko, read 2026-10-15
Governance itself runs on UNI: a holder delegates it to vote, a proposal needs 1,000,000 delegated UNI to reach the floor — lowered from a higher floor by a 2024 vote, down from 2,500,000 on 2024-02-04 — and 40,000,000 UNI must vote for a proposal to pass at all.
Source Uniswap Agora (Fondation Uniswap), read 2026-10-15
The fee switch that funds the burn above pays liquidity providers first: on Uniswap v2, the 0.30% swap fee now splits 0.25% to liquidity providers and 0.05% to the protocol, down from the full fee paid entirely to providers before 29 December 2025.
Source Uniswap Agora (Fondation Uniswap), read 2026-10-15
- What the token is forDirectly verified
- How many existDirectly verified
- How new ones appearDirectly verified
- Who received themDirectly verified
- Who can change the rulesDirectly verified
- Which contract is the official oneThird-party supported
The KLEAR3 rating
3.3out of 5 in the KLEAR3 rating
Each axis gets 0 to 5 from the facts in the register, against thresholds written in advance. The overall score is their weighted average, and it only appears once at least 5 of the 6 axes rest on a measured fact.
- Audits ×1.3
- Open code ×1.2
- Business model ×1.1
- Token alignment ×1
- Governance ×1
- Maturity ×0.7
- Open code3
public repository · 1 commit in 90 days
What holds it back licence unresolved: readable but not reusable · no published release
- Audits1
2 firms on the core · 1 audited scope
What holds it back no report opened: we do not know what they found
- Business model5
business model published by the protocol · paid by whoever swaps · collected by liquidity providers and, since December 2025, the protocol as well, via TokenJar, on Uniswap v2 and a set of high-volume v3 pools · fees paid to the protocol accumulate in an immutable contract (TokenJar); they can only be spent by burning an equal value of UNI, through a second contract (Firepit)
- Governance2
governed through UNI
- Token alignment5
total supply read on-chain · circulating share known · emission schedule published · team and investors fully vested since 2024-09 · terminal inflation 2% a year
- Maturity5
6 years of public code · 1 network verified by us · 8 networks in total
We rate the protocol: how it is built, how much of it can be checked, how its token lines up. Never an investment. A high score is not advice to buy, and a low one is not advice to sell.
What we could not establish
What the two audit PDFs actually found, finding by finding, is not established: no PDF reader was available this session to open either report. Who exactly decided the fee switch’s exact percentages during the UNIfication vote, beyond the numbers the forum itself publishes, is also not something we traced further.
ABDK and Trail of Bits both reviewed the core pool contract, and Trail of Bits’s own README describes three formally verified properties, all confirmed. What remains unread, by us, is the severity table inside either firm’s PDF: how many critical, high, or medium findings each report raised, and which were fixed before launch.
Source ABDK + Trail of Bits, read 2026-10-15
The v3 fee-tier percentages quoted above came from a press account of the vote, not from the governance forum’s own text, which we could read directly for the v2 numbers and the execution date but which did not spell out the v3 split in the pages we opened.
Source KuCoin, read 2026-10-15
Frequently asked questions
Does Uniswap still pay liquidity providers the whole fee, or does the protocol take a cut now?
Both, since 29 December 2025. Before that date, every swap fee on Uniswap v2 and v3 went entirely to liquidity providers. A governance vote called UNIfication changed that: on v2 the 0.30% fee now splits into 0.25% for liquidity providers and 0.05% for the protocol.
On v3, the split depends on the pool's own fee tier: the protocol keeps 25% of LP revenue on low-fee pools and 16.7% on the highest-fee ones, covering pools that carry most of Ethereum mainnet's trading volume.
Those fees land in TokenJar, an immutable contract; they can only be spent by burning an equal value of UNI through a second contract, Firepit. A retroactive 100,000,000 UNI burn, backdated as if the switch had run since launch, came with the same vote: a one-time gesture, worth close to six hundred million dollars at the time.
Is a “NOASSERTION” licence the same as no licence at all?
Not quite, and the difference matters if you want to build on this code. NOASSERTION is GitHub’s own label for a repository where its detector could not identify a recognised open-source licence; it is not a statement that none exists, and not a grant of any reuse right.
That is why a directory printing a green licence badge here is filling in a blank, not reading one. Anyone can read the code; nobody has been granted the right to copy, modify, or redistribute it, which is the whole point of a licence.
Two firms, ABDK and Trail of Bits, did review the core pool contract, and Trail of Bits formally verified three properties of the code with a model checker, all confirmed. Neither outcome tells a reader what they can legally do with the code itself.
If UNI gets burned, why does the total supply still read a billion?
Because burning does not call a function that lowers the counter: it sends UNI to an address nobody holds the key to. We read the contract ourselves. Its own totalSupply() still returns exactly 1,000,000,000, the figure minted at genesis in 2020, untouched.
We read that address, too: it holds 112 895 581 UNI, moved there since the fee switch went live. CoinGecko reports Uniswap's total supply as roughly 887 million, treating that balance as gone; the contract itself disagrees, by design, and neither party is lying.
Neither number is wrong. One counts what the contract will hand out if asked; the other counts what a holder could ever actually move. A reader comparing two sources that both say “total supply” should ask which of the two they mean, especially before quoting either one to someone else.
Where to read on, at the source — every address answered 200 on 2026-10-15.
Related: our Chainlink profile carries the same NOASSERTION licence tag, for a different reason, our BNB profile covers one of the other networks Uniswap deploys on, and our methodology sets out what counts as a source here.
Ask an assistant to summarise it and check what it claims.