Swap routing infrastructure
Quotes that hold.
Every Nordstern quote is a simulated transaction — across 99 EVM networks, down to the tokens other routers get wrong.
$ curl api.nordstern.finance/aggregator/8453 \
?src=<WETH>&dst=<USDC>&amount=1e18&recipient=0xYou…
{
"minToAmount": "1852567340",
"simulationBlock": 49491594,
"gasEstimate": "1064560",
"tx": { "to": "0xC87D…fC3d", "data": "0x3f0b…" }
}Quote honesty
Tier 1
under 5% overquote
Independent Enso Shield, independent benchmark, Apr 2026 · view methodology
Routed volume
$2.285B
routed, cumulative — self-reported
Self-reported DefiLlama display of our own tracker · how it’s measured
The problem
The number you show is not the number your user gets.
Most routers quote an estimate and hope execution catches up. It rarely does — and when it misses, it only ever misses in one direction. Enso’s independent benchmark of aggregator quoting found that “aggregators consistently overquote expected outcomes relative to what execution actually delivers.” [Independent — Enso, Apr 2026]
One example from that benchmark: a single swap overquoted by 3.2% — roughly what a user might expect to earn over an entire year supplying to Aave. [Independent — Enso] That is not a rounding error. It is a full year of yield, lost to a number that was never real.
The pattern is structural, not accidental: an inflated quote wins the comparison your users see. The gap is paid by whoever trusted the number — your user, and then your support queue.
Comparison to Aave supply yield is Enso’s own framing, not Nordstern’s estimate. [Independent — Enso, Apr 2026, blog.enso.build]
The mechanism
We don’t estimate. We simulate.
A Nordstern quote is not a prediction of what a swap will do. It is the result of actually doing it — against a copy of the chain, before you ever send a transaction.
What happens on every request:
- We construct the full transaction — every hop, every call, real calldata.
- We recreate the relevant chain state at the moment of the request.
- We simulate the exact execution path against that state.
- We return the simulated output as the quote. Not an estimate of it. The output itself.
Estimated quoting
- Prices a route from pool state and a formula
- Assumes standard token transfer semantics
- Fails silently at execution when the assumption breaks
- Wins comparisons; loses execution
Simulation-first quoting
- Executes the route against real, current chain state
- Handles fee-on-transfer, rebasing and taxed tokens because it actually runs the transfer
- Fails at quote time, before you ever show the user a number
- Wins comparisons and execution, because they’re the same computation
The invariant: our quote is not an estimate. It is a simulation of the exact transaction we hand you. Any difference at execution comes from the market moving in the seconds afterward — never from an optimistic guess.
The proof
Where quote integrity meets coverage
The accurate routers are narrow. The broad routers are the worst offenders. It’s the one quadrant that isn’t empty.
Chart data — every plotted position, with sources
| Vendor | Enso tier | Named figure | Vendor-published coverage |
|---|---|---|---|
| 0x | Tier 1 | 0.25% overquote | 21 chains, ~45M tokens |
| SushiSwap | Tier 1 | named standout on simulation failure | narrow — count not compiled here |
| Nordstern | Tier 1 — under 5% overquote | — | 99 EVM networks, 1,000+ DEXs, 465,000+ pools, 10M+ indexed tokens |
| LI.FI | Tier 3 — “inflation is essentially universal” | — | broad, multi-chain |
| OpenOcean | Tier 3 | 100% overquote | broad, multi-chain |
| Not graded by Enso Shield, therefore not plotted: Bebop, among others — placing an ungraded vendor on a measured axis would be an unsourced claim. | |||
Nordstern’s quadrant — high integrity, broad coverage — is close to empty. One honest exception, and one qualification worth stating plainly: 0x sits above us on integrity. Its published overquote figure, 0.25%, is better than ours — no figure on the board beats it. We’re not going to argue otherwise; the board is public.
What 0x doesn’t cover is the reason the gap doesn’t close the argument. 0x’s accuracy is measured across 21 chains and roughly 45M tokens — the liquid, well-behaved part of the market. Nordstern holds Enso’s Tier 1 band across 99 EVM networks, 10M+ indexed tokens, and the fee-on-transfer and rebasing tokens that break routers built for standard transfers. Being accurate on ETH→USDC is table stakes. The harder, and more expensive, thing to be accurate on is everything else — and that’s the part of the market where a broken quote actually costs you a support ticket.
So: if your flow is entirely liquid, mainstream pairs on a handful of chains, 0x’s number is the better one to quote against. If your flow includes the long tail — new tokens, taxed tokens, chains outside the top 20 — query both and look at your own results. That’s not a hedge. It’s the test we’d run in your position.
The long tail
Coverage isn’t the pitch. It’s the difficulty setting.
Anyone can be accurate on ETH→USDC. The pools are deep, the tokens behave exactly as the ERC-20 standard says they will, and every router on the market gets it right. Integrity is easy where liquidity is easy.
It gets hard in the long tail — and the long tail is where your users increasingly trade:
- Fee-on-transfer tokens, where the amount that arrives is not the amount that was sent. A router that assumes standard transfer semantics quotes the wrong number by construction.
- Rebasing tokens, where balances change between quote and settlement for reasons that have nothing to do with the trade.
- New and thin tokens, where a pricing formula extrapolates from too little data and a simulation either works or reverts — no in between.
We don’t handle these as a separate feature. They fall out of the same mechanism: if you actually construct and simulate the transaction, taxed-token accounting and rebasing balances are just what the simulation returns. A router that estimates has to special-case every one of these. A router that simulates doesn’t have to know they’re special.
99 EVM networks · 1,000+ DEXs · 465,000+ pools · 10M+ indexed tokens. [Own claim] Read those not as a coverage count, but as the size of the difficulty setting our integrity claim has to survive.
The evidence
Every number here, labelled by who’s vouching for it.
Independent
Nordstern does not control these.
- “Tier 1 (under 5% overquote)” — Enso Shield, the independent DeFi aggregator benchmark, dated 2026-04-22. Verbatim from the source; we haven’t rounded it up. View the methodology
- “Simulation failure comfortably under 0.5%” — Enso Shield, same source and date. (Note, in fairness: Enso names SushiSwap, not Nordstern, as the standout on this metric — we’re in the same band, not the leader of it.)
Partner-observable
Revealed preference of sophisticated buyers who could route elsewhere.
- Integrated by 1delta, Enso, LI.FI, Matcha Meta, Mayan, deBridge and Oku — stated as integration relationships, not as volume shares we haven’t independently verified. Each queries Nordstern as one provider among several. [Company-stated, 2026-08-03; Matcha Meta and Oku publicly observable]
- Yes, that Enso — the benchmark provider cited above also integrates Nordstern. Stated here so you don’t discover it: the two roles are separate. Shield grades 20 aggregators on a published methodology, and the integration has no bearing on the grading — 0x’s better overquote figure on that same board is the proof the referee isn’t doing us favours.
- One of 14 swap routers on Oku’s public docs. View Oku docs
- Live on Matcha Meta since July 2026 — announced publicly by both sides. [Partner-observable]
1delta · queries Nordstern as one provider among several
LI.FI · queries Nordstern as one provider among several
Matcha Meta · live since Jul 2026
Mayan · queries Nordstern as one provider among several
deBridge · queries Nordstern as one provider among several
Enso · queries Nordstern as one provider among several · also runs the benchmark we cite — disclosed above
Nordstern telemetry
Self-measured — labelled as such, always with conditions.
- $2.285B cumulative routed volume; more than $1B in the trailing 12 months — displayed on DefiLlama, but its adapter calls our own volume tracker rather than reading the chain, so this is self-reported. The mechanism, in full
- <50ms average quote latency, measured from our own datacenters. Enso’s benchmark, which measures from US East, records us around 90ms from that vantage point — both numbers are accurate; they measure different distances.
- 1k+ quotes/second sustained throughput.
- Underquote rate — quoting below what execution delivers — measured and published with its conditions stated, not as an unconditioned headline figure.
Architectural
Weak as a claim. Strong as an explanation. Instantly checkable.
The mechanism above, and the simulationBlock in the response you get back below — the API is its own evidence.
Who it’s for
Four kinds of teams route through Nordstern.
Meta-aggregators & bridges
Your quote is only as honest as your providers. We’re independently graded Tier 1; add us to the set and see what it does to yours.
Perps, lending, vaults & treasuries
Convert arbitrary collateral without a failed transaction — including the fee-on-transfer and rebasing tokens that break naive routers.
Wallets & swap widgets
Show your users a number you can stand behind. Fewer reverts, tighter slippage tolerances, long-tail coverage your support team doesn’t hear about.
Bots, searchers & market makers (EVM)
Deterministic, simulated output — <50ms average quote latency, no API key required. Colocation on request.
Integration
No API key. No contract. One request.
curl "https://api.nordstern.finance/aggregator/8453?\
src=0x4200000000000000000000000000000000000006&\
dst=0x833589fCD6eDb6E08f4c7C32D4f71b54bdA02913&\
amount=1000000000000000000&\
recipient=0xYourWallet..."{
"src": "0x4200000000000000000000000000000000000006",
"dst": "0x833589fCD6eDb6E08f4c7C32D4f71b54bdA02913",
"fromAmount": "1000000000000000000",
"minToAmount": "1852567340",
"simulationBlock": 49491594,
"gasEstimate": "1064560",
"tx": {
"from": "0xYourWallet...",
"to": "0xC87De04e2EC1F4282dFF2933A2D58199f688fC3d",
"value": "0",
"data": "0x3f0bde25…"
}
}Base (chain id 8453) · simulationBlock 49491594 · captured 2026-08-03 · tx.data truncated, wallet address substituted — otherwise verbatim from the live API
Three fields in this response are the argument:
minToAmount — not an estimate of what you’ll get; the floor of what you’ll get. Execution is capped at the quote, so the field name encodes the promise: this number or better, never less.
simulationBlock — the exact chain state this quote was simulated against. We publish it on every response so anyone — you, or an independent benchmark — can re-simulate at that block and check us. A quote you can’t verify is a claim; this one is a measurement with its conditions attached.
gasEstimate — measured from the simulation, not a heuristic.
Pricing and economics
Free to integrate. You keep 100% of your own fees.
No API key to start. No contract to sign before your first quote. If you add a fee on top of ours, it’s entirely yours — we don’t take a cut of it. [Own claim]
How we actually earn, in our own words, before you ask:
Execution is capped at the price we quote you. Your user never receives less than the number we showed. If the trade settles better than that — market movement in Nordstern’s favour between quote and execution — that improvement accrues to Nordstern, not to the end user.
That’s a real trade-off, and it’s worth being direct about it: it means the tighter and more honest our quote is, the less room there is for that improvement, and the less we earn per trade. We only make money by beating our own promise, never by missing it. Quoting honestly is structurally against our own short-term interest — which is exactly why you can believe the quote we give you.
One more number worth publishing precisely because it’s the obvious place to be suspicious: our underquote rate — quoting below what execution actually delivers — is measured and published with its conditions stated. [Nordstern telemetry] A model that only ever benefits from tight quotes would show up here first if it were drifting toward conservative underquoting; we publish this number so that drift would be visible before it became a pattern.
Next step
Add Nordstern to your routing set.
Not a migration. Not a contract. Query us alongside whatever you already run, and look at your own numbers — win rate, revert rate, how often our quote matched settlement. If we don’t win quotes on your traffic, we cost you nothing.
Talk to us covers colocation siting, a network we haven’t indexed yet, and volume-tier terms.