Tokenised equities trade in dozens of pools that disagree with each other. Keel reads them as one book, quotes every venue at the size you are actually trading, and reverts any fill a single venue would have beaten.
Quotes read live chain state at the moment you ask. The router is unaudited and the best-execution proof covers registered venues only.
The problem
Sorting a book by price is what “best bid and offer” means. Here is NVDA against USDG sorted exactly that way — beside what each venue can absorb before it moves 1%, and what $10,000 would actually get you there.
| Venue | Advertised price | Depth to 1% | $10,000 fills |
|---|---|---|---|
| v4 · 0.008 %not routable | $ 225.1531 | <$1 | 0.000660 sh |
| v4 · 0.300 % | $ 225.3761 | $35,826 | 44.154 sh |
| v4 · 1.000 %not routable | $ 225.4018 | $38 | 0.5553 sh |
| v4 · 0.037 %not routable | $ 225.7051 | $509 | 40.333 sh |
| v3 · 0.050 % | $ 225.7977 | $513,112 | 44.261 sh |
| v3 · 0.300 %not routable | $ 226.0235 | $2,368 | 42.926 sh |
The venue marked not routable advertises the best price in this book on $0 of depth, and fills 0.000660 shares. Any router that sorts by price sends your whole order into it. This book shows every venue on the chain — including the ones we refuse to trade on, because hiding them would hide what you are being protected from.
Read from chain state at block 53,253,320 on 2026-09-03 07:48 UTC, 88h ago. This table is a snapshot and does not move on its own. The quote in the panel above is not — it calls the router on-chain the moment you ask for it.
What it costs
67,042× apart. The venue on the left is the one advertising 29 basis points cheaper — it works out at $15,146,944 a share against $225.93. Both figures come from the same quoter the router runs on-chain, against live state at block 53,253,320. Keel quotes every venue and routes to the one that can pay.
1,797 v4 pools touch AAPL across 101 hook contracts, and four contracts answer to the name “NVIDIA · Robinhood Token”. Exactly one of them is NVIDIA. Assets are identified by ABI rather than by name, and a venue has to be unhooked, priced under a 1% fee, and deep enough to absorb $2,500 before it is eligible for an order.
Spot price answers a question nobody is asking. Keel walks each venue’s liquidity curve tick by tick at the size you are actually trading — the same walk the pool performs when it fills you. The walk exists twice, once in TypeScript and once in Solidity, and the two agree to the wei.
One venue, or several when splitting genuinely earns something. At retail size it usually does not, because the cheapest venue is normally also the deepest — so the solver picks a single place and says so, rather than carving the order into legs for the look of the thing.
Before a single leg executes, the contract re-quotes every registered venue at your size. After execution it compares the fill against that benchmark. If any one venue would have done better, the transaction reverts and the money stays where it was.
Across the chain
Every indexed asset quoted on more than one venue, widest disagreement first. The megacaps are tight. The gaps are in the names nobody is watching.
| Asset | Venues | Cheapest | Dearest | Spread | Deepest venue |
|---|---|---|---|---|---|
| SPCX | 4 | $ 140.86 | $ 141.99 | 80 bps | $ 286,577 |
 COIN | 2 | $ 176.67 | $ 177.88 | 68 bps | $ 13,222 |
 AMD | 2 | $ 455.53 | $ 458.27 | 60 bps | $ 18,330 |
 NFLX | 3 | $ 82.72 | $ 83.22 | 60 bps | $ 8,719 |
 PLTR | 2 | $ 169.27 | $ 170.16 | 53 bps | $ 68,417 |
 INTC | 3 | $ 90.16 | $ 90.61 | 50 bps | $ 3,108 |
 META | 3 | $ 595.46 | $ 598.24 | 47 bps | $ 18,847 |
 AAPL | 4 | $ 324.12 | $ 325.46 | 41 bps | $ 159,160 |
 MSTR | 2 | $ 124.65 | $ 125.14 | 40 bps | $ 28,490 |
 NVDA | 6 | $ 225.15 | $ 226.02 | 39 bps | $ 513,112 |
 SLV | 2 | $ 59.46 | $ 59.66 | 33 bps | $ 18,547 |
 GOOGL | 4 | $ 338.36 | $ 338.97 | 18 bps | $ 70,949 |
 GLD | 4 | $ 405.10 | $ 405.76 | 16 bps | $ 318,382 |
 SPY | 4 | $ 766.20 | $ 767.37 | 15 bps | $ 3,398,279 |
 QQQ | 4 | $ 709.50 | $ 710.20 | 10 bps | $ 227,878 |
 TSLA | 2 | $ 359.08 | $ 359.37 | 8 bps | $ 36,741 |
 MSFT | 3 | $ 497.81 | $ 498.16 | 7 bps | $ 48,170 |
 AMZN | 2 | $ 256.21 | $ 256.37 | 6 bps | $ 37,050 |
FAQ
A displayed price is the price for an infinitely small trade. The moment you buy you move it, and how far you move it depends entirely on how much money sits in the pool — which the price does not tell you.
The venues in the book above quote within a few basis points of one another while differing by orders of magnitude in what they can absorb. Price sorts them. Depth decides them.
Precisely this: no single registered venue could have filled your order better. The contract quotes them all at your size before executing, then checks the result against that benchmark. If it was beaten, it reverts.
It is deliberately not a claim of global optimality across every possible split — that is an exponential search. It is the claim you care about, and the one a contract can actually prove inside a transaction.
No. One transaction: tokens enter, every leg executes inside a single Uniswap v4 unlock, and the output lands with you — or the whole thing reverts and nothing moved.
The router holds nothing at rest, and a test asserts it retains no balance after any fill.
Usually not, and we would rather say so than sell the feature. On most assets at retail size the split solver earns nothing at all, because the cheapest venue is nearly always also the deepest.
It is not always zero. On the more fragmented names, once an order is large enough to start moving the best venue, the solver does find a basis point or two — we have measured low single digits on AAPL at $25,000. The panel reports the gain it actually found on your size, including when that gain is nothing.
The proof covers venues in the registry. If a pool is not on that list, nothing is proven about it — and the list is curated, so you are trusting that curation.
The contract is unaudited. The owner can add and remove venues; entries are validated against the canonical Uniswap factory, so a compromised owner cannot point your order at their own contract, but they can still choose which real pools are routable.
Liquidity concentrates where attention does. Megacaps have deep pools that arbitrage keeps aligned. Thin names have shallow pools nobody bothers to arbitrage — and that is exactly where a fill goes wrong quietly.
One pool. A Uniswap v3 pool at a specific fee tier, or a v4 pool identified by its key. The same pair commonly exists at four fee tiers across both protocols, and each one is a separate market with its own price and its own depth.