Is NEAR becoming a cross-chain settlement layer via NEAR Intents? A three-hour sample of unauthenticated dry quotes, set against public Kraken mids.
An intent is the outcome, not the route. NEAR Intents asks a market of solvers to find a way from the asset you hold to the asset you want, including across chains. That is a settlement claim only if the quotes exist and the price is close to a market a person can already see.
Three hours of dry: true 1Click quotes cannot show that NEAR is becoming that layer. They can show whether, in this window, the quote was usually there, and whether the token amount sat near a Kraken cross-mid. A third, descriptive check asks whether DefiLlama’s published daily volume is expanding. The pre-registered bars are in PREREGISTRATION.md.
Each cycle posts to https://1click.chaindefuser.com/v0/quote with dry hard-coded true and no API key. The published fee schedule says an unauthenticated quote carries an extra 25 basis points. That fee stays inside the measured gap. Adding 25 basis points back is labeled an estimate and is not the test.
Ten routes, two sizes: NEAR and USDC both ways, USDC from Ethereum to Solana and to Base, USDT from Ethereum to Tron, ETH to SOL, BTC to ETH, SOL to NEAR, ETH from Arbitrum to Ethereum, and NEAR to SOL. That is seven chains: NEAR, Ethereum, Solana, Bitcoin, Tron, Base, and Arbitrum. Stable and same-asset routes are one family. One extra 10 wNEAR quote waits 2,000 milliseconds and is left out of the rates and the medians.
The price is whole tokens from the integer amounts. Kraken’s mid is the average of the best bid and the best ask. The cross is the origin mid divided by the destination mid. OKX and Coinbase are stored and shown. They do not decide the pass. Binance’s book ticker returned HTTP 451 from this host and is not a reference.
A cycle counts when all 20 standard quotes were attempted. Public volume is DefiLlama’s daily chart and, separately, the volume block on the Intents explorer homepage. The explorer API itself answers MISSING_TOKEN without a JWT and is not used for totals. The two series are never averaged.
Every figure is drawn from data/verdict.json and data/series.json, which analysis/rebuild.py writes from data/raw. Hover a chart. The route row under figure 3 switches the series.
A miss is a timeout at 50 seconds, a Cloudflare 520, or a truncated body. Those eight rows stay in the raw file. They are not dropped from the rate.
The route median in the table sits between these two sizes. It is not the price of either size. The official volatile test pools both with the other volatile routes.
Raw rows live in data/raw. The dataset manifest is anchored on NEAR testnet. The anchor shows the bytes were not edited later. It does not make a quote true.
| Route | Family | Quotes | Kraken | OKX | 1Click USD |
|---|
Availability is falsified below 50 percent, or if fewer than five routes ever quote. It passes only at or above 85 percent, with at least eight of the ten routes at or above 70 percent. The price statement passes only if the stable median is at least −40 basis points and the volatile median at least −100. It is falsified past −100 and −200. Between those bars the result is “not supported.”
A dry quote is not a settled transfer. Comparing with a mid makes Intents look worse than a price you could lift on a central book, and this study does not add central-book fees either. The gap is mid against mid. 1Click’s own dollar marks sometimes disagree with the token amounts, which is why they are not the official test. Same-asset routes are closer to a bridge quote than to price discovery.
DefiLlama’s 24-hour total and the explorer homepage have matched at some fetches. The 7-day, 30-day, and all-time figures have not, and DefiLlama’s total30d headline is not the sum of the last 30 daily bars. growth_30 uses the bars. The newest bar in the snapshot is dated before the sample day. Rising volume would still not prove the strong claim. Contracting volume, which is the label this window gets, does not by itself erase a working quote.
The clock span is one host and one network path. Quotes move; re-running the collector will not reproduce these rows. Re-running python3 analysis/rebuild.py will reproduce the figures from the committed rows. No swap was submitted.