Whitepaper

A non-custodial cross-chain swap protocol with native Bitcoin support. Version 1.0.

Abstract

Cross-chain value transfer is the most attacked surface in cryptoeconomics. Approximately $2.5 billion has been lost to bridge exploits since 2022, distributed across Wormhole, Ronin, Nomad, Multichain, Harmony, and Orbit. Each incident shares the same architectural pattern: tokens locked on a source chain, wrapped representations minted on a destination chain, and a custodian — a multi-signature scheme, MPC validator set, or optimistic light client — guarding the locked treasury. The locked treasury is the attack surface.

Flip is a non-custodial cross-chain swap protocol that does not use this pattern. Orders settle through NEAR Intents, a solver network where independent market-makers post collateral and compete to fill cross-chain swaps. Solvers that fail to deliver are slashed and users are refunded automatically. No locked treasury, no custodian, no wrapped IOUs. Native Bitcoin — not WBTC, cbBTC, or BTCB — enters and exits the system on the actual Bitcoin network.

Six chains are supported at launch: Bitcoin, Ethereum, Solana, NEAR, Arbitrum, and Base. Measured end-to-end settlement on production code paths is under sixty seconds on layer-2 corridors, with effective all-in cost of 0.07–0.21% on swaps benchmarked against equivalent centralized-exchange execution at approximately 47% lower cost. The protocol fee is 0.5%, displayed in every quote before execution, with no off-quote surcharges. The operating entity is Flip Pty Ltd, an Australian limited company; sanctions screening is performed server-side against the Chainalysis OFAC database with a circuit breaker that fails closed on sustained screening outages.

1. Introduction

1.1 The problem

A user holding Bitcoin on the Bitcoin network and wanting to receive USDC on Arbitrum has three options under current cross-chain infrastructure. None are good.

The bridge. Lock BTC in a vault on Bitcoin, mint wrapped BTC on Ethereum, swap on a DEX, bridge to Arbitrum, swap again. Two bridge legs, two DEX legs, six minutes of latency in the optimistic case. The cumulative attack surface is the locked Bitcoin treasury — historically the asset class most aggressively pursued by adversaries. Every major lock-and-mint bridge has been compromised at least once.

The centralized exchange. Deposit BTC, sell for USD, withdraw USDC to Arbitrum. Resolves the security question by accepting custody. Costs: KYC requirement, deposit-to-withdrawal latency of fifteen to ninety minutes, and the requirement that the exchange operates in the user's jurisdiction. For users in Australia, the Asia-Pacific time zone exacerbates the latency: an exchange withdrawal queued at 9 a.m. Sydney time may not process until US business hours.

Wrapped Bitcoin variants. Hold WBTC, cbBTC, or BTCB and skip the bridge entirely. Resolves latency and avoids the bridge attack surface but introduces a different problem: these tokens are structurally IOUs. Each one is backed one-to-one by Bitcoin held by an identifiable custodian (BitGo for WBTC, Coinbase for cbBTC, Binance for BTCB). The trust assumption has not been eliminated; it has been displaced from the bridge operator to the wrap operator. For Bitcoin holders whose original choice to hold BTC was a statement about non-custodial ownership, this trade is unacceptable.

None of these paths preserves the property the user actually wants: the destination asset arrives on the destination chain, the source asset is delivered through native infrastructure, and no third party holds the user's funds at any point.

1.2 Existing approaches and their failure modes

The cross-chain category has converged on three architectural patterns, each with documented failure modes.

Lock-and-mint bridges. Lock the source asset on chain A, issue a wrapped representation on chain B. The locked treasury must be guarded by something — multi-signature, MPC, optimistic light client, or zk-proof. Documented failure modes: signature collusion (Ronin: $625M), replay vulnerabilities (Nomad: $190M), guardian-key compromise (Harmony: $100M), oracle failure (Multichain: $125M). The aggregate loss since 2022 is approximately $2.5 billion.

LP-based liquidity networks (THORChain, Maya). Solvers or LPs hold inventory on each supported chain; cross-chain swaps consume that inventory directly. The protocol holds inventory, which means the protocol holds attack surface. THORChain has been exploited multiple times. LPs face permanent impermanent-loss across volatile cross-chain corridors, leading to capital flight during turbulence.

Hashed time-locked contracts (HTLCs, atomic swaps). Cryptographically guaranteed atomic execution between two chains. Failure mode is the practical one: poor liquidity, multi-confirmation latency, and unidirectional channel structure that limits real UX to specialist users.

1.3 Contribution

Flip's contribution is not the discovery of intent-based settlement — that primitive exists upstream in NEAR Intents, in parallel form in CoW Protocol's coincidence-of-wants engine, and in UniswapX's filler network. The primitive replaces "the protocol holds inventory" with "solvers post collateral and compete to fill orders." Failure to deliver is slashed; the user is refunded automatically from the protocol, not from a treasury.

Flip's contribution is the consumer-grade application of that primitive to native Bitcoin specifically, alongside five other production chains, delivered through an editorial interface that exposes the full cost structure of every quote before execution and proves, on-chain, that the user's funds are not custodied at any step.

The architectural argument is that the cross-chain category does not require new cryptography. It requires a refusal to ship lock-and-mint and a willingness to publish measurable cost data alongside every transaction.

2. Background: intent-based settlement

2.1 The solver-network primitive

An intent is a signed statement of a desired end-state, not a procedural transaction. The user signs: "I will pay 0.001 BTC; in exchange I want at least 88.62 USDC on Base, delivered to address 0x6F…2349, by deadline T." A network of solvers — independent agents who have posted collateral to the protocol — observe pending intents and compete to fulfill them. The first solver to deliver the destination amount to the destination address inside the deadline wins the trade and is paid out from the user's deposit.

Solvers who accept an intent and fail to deliver are slashed: their posted collateral is forfeited and the user is refunded automatically from the protocol. The economic argument is straightforward — as long as collateral exceeds expected upside from non-delivery (including possible MEV), rational solvers deliver.

The locked treasury of the lock-and-mint pattern does not exist in this design. There is no aggregate pool of user funds for an attacker to drain. Each swap is an isolated contract between user and a specific solver, mediated by the protocol's slashing condition. The blast radius of any single failure is bounded by that solver's collateral, not by the protocol's TVL.

2.2 NEAR Intents as settlement layer

Flip's settlement layer is NEAR Intents, deployed on NEAR mainnet. NEAR was selected over competing settlement layers (UniswapX, CoW Protocol, CCTP) for four reasons.

Native non-EVM and EVM solver participation. NEAR Intents accepts solvers on Bitcoin, Solana, NEAR, and every EVM chain. UniswapX is EVM-only; CoW Protocol is EVM-only. Native Bitcoin participation is a hard requirement for Flip's product thesis and rules out the EVM-only systems.

Active solver competition. As of May 2026, NEAR Intents shows active solver presence on every supported Flip corridor at swap sizes from $1 to $10,000. Solver-side liquidity is the binding constraint on intent-based systems; on Flip's corridors it is currently non-binding.

Public Explorer with verifiable execution. Every settlement is observable on explorer.near-intents.org. Flip's /api/network-swaps and /stats routes proxy this data so any user can inspect the protocol's actual throughput.

Integrator fee primitive. NEAR Intents exposes an appFees parameter on quote requests that lets integrators take a configured cut of each swap. The fee is delivered to the integrator's NEAR account on settlement, in the destination asset's NEAR-wrapped representation, with no off-chain reconciliation required.

2.3 Comparison to lock-and-mint bridges

The last row is not yet a stable claim — intent-based settlement is a younger pattern with a smaller aggregate volume than the bridge category at its peak. The economic security model is, however, structurally different, and the absence of a locked treasury removes the dominant attack class.

3. Architecture

3.1 System overview

Flip is a thin client over NEAR Intents. It contributes:

• A web interface at flipdex.co that constructs intent payloads from user input, displays signed quotes returned by 1Click, and exposes the full quote structure (input, output, slippage, fees, deadline) before execution.
• A server-side proxy at /api/quote that adds compliance enrichment (sanctions screening), enforces the protocol fee, sanitizes the request body, and forwards to NEAR Intents 1Click.
• A status surface at /status and /api/status that lets users (and journalists, and auditors) verify the on-chain execution of any historical swap by deposit address.
• A network-activity surface at /api/network-swaps proxying live NEAR Intents Explorer data, demonstrating that the underlying protocol is operating at scale across all supported corridors.

Flip does not hold inventory, custody, or sign transactions on the user's behalf. The user signs every state-changing transaction with their own wallet.

3.2 The quote lifecycle

A complete swap is structured as four phases.

Compose. The user selects a source asset and chain, a destination asset and chain, an input amount, and a recipient address. The client constructs an intent payload conforming to the NEAR Intents 1Click API schema and POSTs to /api/quote.

Validate. The server validates the intent against compliance and structural constraints. Sanctions screening is performed on the recipient and refund-to addresses against the Chainalysis OFAC database. The request body is sanitized — any client-supplied appFees parameter is stripped to prevent fee-redirection attacks — and the server-configured appFees is injected. The cleaned request is forwarded to 1Click.

Sign. 1Click returns a signed quote containing a deterministic deposit address, the guaranteed output amount, the slippage tolerance, the solver fee, the protocol fee, and the deadline. Flip displays this signed quote to the user. If the user accepts, Flip signs a deposit transaction on the source chain to the deposit address.

Settle. Solvers monitoring NEAR Intents observe the deposit, claim the trade, and deliver the destination amount to the recipient address on the destination chain. The deposit is released to the winning solver. The integrator fee (Flip's 50 bps) is delivered to the configured NEAR account. If no solver delivers inside the deadline, the deposit is refunded automatically.

Each phase is independently inspectable. The user can verify the deposit on the source chain explorer (Etherscan, Basescan, Blockchain.info), the settlement on NEAR Intents Explorer, and the destination delivery on the destination chain explorer.

3.3 Deposit and settlement

Deposit addresses are derived deterministically from the quote payload by 1Click and are time-bound to the quote's deadline. A deposit landing after the deadline is refunded; a deposit landing before is accepted into the solver auction.

Settlement is asynchronous. The user's client polls /api/status?depositAddress=… for state transitions: WAITING_FOR_DEPOSIT → DEPOSIT_DETECTED → SETTLING → COMPLETE or REFUNDED. State is sourced from the NEAR Intents protocol's own state machine; Flip's status route is a proxy and does not maintain independent state.

Production measurement (Section 6) indicates typical L2-to-L2 settlement under 60 seconds end-to-end, with the slow phase being source-chain deposit confirmation (typically 12–30 seconds on Ethereum mainnet, 1–3 seconds on Base/Arbitrum).

3.4 Refund path and slashing

If a solver claims a trade and fails to deliver inside the deadline, the protocol's slashing condition triggers. The solver's collateral is forfeited. The user's deposit is refunded to the configured refundTo address — typically the same address that originated the deposit. From the user's perspective the worst-case outcome is a latency penalty (the deadline window) plus the source-chain gas spent on the deposit transaction, but no principal loss.

This property — that the worst case is a temporal cost, not a financial one — is the qualitative distinction from the bridge category. Bridge users who experienced the Ronin hack lost principal. Flip users in the equivalent failure mode lose deadline-window time.

4. Economic model

4.1 Fee structure

The protocol fee is 50 basis points (0.5%) of the swap value, configured server-side via the FLIP_FEE_BPS environment variable and injected into the 1Click appFees parameter on every non-dry-run request. The fee is displayed in the quote summary prior to user confirmation, alongside the solver fee and destination network fee. There are no off-quote surcharges.

The fee is bounded by code to the range 0–500 bps (0–5%) with the upper bound enforced at the server-side fee-injection logic. The fee can be disabled per-deployment by setting FLIP_FEE_BPS=0.

Dry-run quotes (used for display-only previews on the compose screen) do not have fees injected. This is deliberate: dry-run quotes are intended to surface the protocol's underlying cost structure before the integrator overhead is applied, allowing users to verify the fee is the only Flip-side cost added to the trade.

4.2 Fee distribution mechanism

Fee distribution is performed entirely by the NEAR Intents protocol on settlement. Flip does not participate in fee collection beyond specifying the appFees parameter at quote time. The fee is delivered to the configured NEAR account (FLIP_FEE_RECIPIENT) as a NEAR-wrapped representation of the source or destination asset, depending on settlement path.

Empirically, fee delivery is observed within 30–60 seconds of swap completion, denominated in the destination corridor's settlement asset. The Flip operating account on NEAR mainnet is largevinegar3245.near and is publicly verifiable at nearblocks.io.

4.3 Measured take rate

Two production swaps were executed against mainnet code paths during initial validation:

• Swap 1: 0.001 ETH (mainnet) → 0.000999 ETH (Base). Input $2.12, output delivered to Base; fee delivered to largevinegar3245.near as 0.000003 ETH (~$0.0064). Observed effective fee rate: ~0.30%.
• Swap 2: 0.000594 ETH (Base) → 1.2527 USDC (NEAR). Input $1.26, output delivered to NEAR; fee delivered as ~0.000001 ETH (~$0.002). Observed effective fee rate: ~0.17%.

The discrepancy between the configured 50 bps and the observed 17–30 bps is attributable to (a) NEAR-wrap denomination spread at sub-cent values introducing display-precision artifacts, and (b) likely intra-protocol distribution mechanics at NEAR Intents that may differ from the documented 50 bps headline rate at small swap sizes. Definitive take-rate measurement requires a swap volume above $30–50 where fee values are larger than wrap-spread artifacts. This is a near-term measurement priority.

5. Trust and security

5.1 Non-custodial property

At no point in the swap lifecycle does Flip, NEAR Intents, or any third party hold the user's funds in a manner the user has not explicitly authorized for a specific transaction. The deposit transaction sends funds to a deterministic, time-bound address controlled by the NEAR Intents protocol. The destination asset is delivered directly from the solver to the recipient address. The integrator fee is delivered from the protocol settlement, not deducted from a Flip-controlled account.

The non-custodial property is verifiable post-execution by reading on-chain transaction data. See /trust for the step-by-step verification protocol.

5.2 Web surface defenses

The application surface defends against three attack classes. Clickjacking and frame embedding via X-Frame-Options: DENY and frame-ancestors 'none' on every route except the explicit partner-iframable surface at /embed. Cross-site script injection via a strict, enforcing Content-Security-Policy that allowlists only the specific external hosts the application contacts at runtime. Transport-layer downgrade via HSTS with two-year max-age, includeSubDomains, and preload.

5.3 Compliance layer (sanctions screening)

Recipient and refund-to addresses are screened against the Chainalysis OFAC sanctions database server-side at quote time. The screening is performed via the Chainalysis Free Sanctions Screening API with a 3-second request timeout and 24-hour result cache.

The screening operates under a circuit-breaker pattern: after five consecutive Chainalysis API errors, the breaker opens and all subsequent quote requests are rejected with HTTP 503 for a five-minute cooling period. This is the fail-closed posture: during sustained screening unavailability, swaps are refused rather than proceeding without verification.

5.4 Operational risks

Residual operational risks, ordered by likelihood: phishing and lookalike-domain attacks (defensive registration maintained); solver liveness (upstream NEAR Intents dependency, currently non-binding); browser extension and wallet compromise (wallet- layer risk, mitigated by user verification of recipient address in the wallet's own confirmation dialog); NEAR Intents protocol bug (upstream dependency, mitigated by NEAR Intents being open-source, audited, and in production at meaningful volume).

6. Empirical performance

6.1 Methodology

Performance numbers cited in this paper are measured from production code paths against mainnet, not from staging or simulation. Each swap was executed end-to-end using the same client code that ships to users at flipdex.co. All measurements are independently verifiable on public block explorers.

6.2 Measured settlement times and costs

ETH (Ethereum) → ETH (Base) corridor. Deposit transaction 0xbf83… confirmed on Ethereum mainnet at block 25164521 with gas cost $0.09. Settlement on NEAR via execute_intents on intents.near completed in approximately 47 seconds from deposit confirmation. Quoted effective cost: -0.07%.

ETH (Base) → USDC (NEAR) corridor. Deposit on Base with gas cost under $0.001. Settlement on NEAR in approximately 35 seconds. Destination delivery as USDC (NEP-141 token on NEAR) to largevinegar3245.near. Quoted effective cost: -0.21%.

ETH (Ethereum) → USDC (Ethereum, same-chain) corridor. Quoted effective cost: -14.18%. This corridor performs poorly because the solver must absorb mainnet gas on both sides of a small swap, and the structural mainnet-gas overhead dominates trade economics below ~$50 swap size. Flip surfaces this with an inline warning before execution.

The honest cost disclosure on the same-chain mainnet corridor is itself a product feature: every quote exposes its true economics, and the application actively suggests alternative routes when the user is on a structurally expensive one.

6.3 Comparison to centralized and bridged alternatives

The Flip quote engine computes side-by-side cost comparisons against Coinbase and against a generic lock-and-mint bridge route for each quote, and displays the savings. For the ETH (Ethereum) → ETH (Base) corridor at $2.12 swap size: Coinbase-equivalent cost was approximately $1.01 higher than Flip (47.79% more expensive). Bridge-equivalent cost was $1.51 higher (71.44%). For the ETH (Base) → USDC (NEAR) corridor at $1.26: Coinbase $0.51 higher (40.07%), bridge $1.01 higher (79.76%).

At $50 swap size on the same corridors, the percentage savings remain stable while the absolute savings scale linearly. At $500, the absolute savings exceed $20 per swap; at $5,000, exceed $200.

7. Operational posture

7.1 Jurisdiction and legal structure

Flip Pty Ltd is an Australian limited company headquartered in Sydney. Australian operation is load-bearing for three reasons.

Regulatory clarity. AUSTRAC has issued specific guidance distinguishing non-custodial swap aggregation (which Flip performs) from custodial exchange services (which Flip does not), with proportionate compliance obligations.

R&D Tax Incentive. Australia's R&D Tax Incentive provides a tax-offset claim for eligible R&D expenditure. Flip's development qualifies under the eligible-activity criteria.

Time zone. Asia-Pacific operation provides a structural alpha for solver- side participation in NEAR Intents' continuous markets and a structural alpha for inbound investor diligence from US-coast and European VC processes — meetings can be scheduled in either party's morning.

7.2 Verifiability

Every claim in this paper is testable. The protocol fee can be inspected in src/app/api/quote/route.ts. The sanctions screening logic and circuit breaker is in src/lib/sanctions.ts. The CSP and security headers are in next.config.ts. The NEAR Intents Explorer at explorer.near-intents.org confirms every settlement.

There is no off-chain ledger, no closed-source matching engine, no opaque settlement layer. A diligence visitor can verify every protocol claim made above in approximately fifteen minutes using only public block explorers.

8. Limitations and future work

Solver liveness risk. The protocol's user-visible reliability is a direct function of NEAR Intents' solver-network liveness. Flip cannot directly mitigate solver liveness; the mitigation is operational monitoring of the upstream protocol.

Mobile EVM wallet connection. Currently restricted to wallet-in-app browsers (MetaMask Mobile, Coinbase Wallet, Rabby) because the application uses the injected()wagmi connector without WalletConnect. The mobile EVM gap is signposted in the wallet picker on mobile devices ("Open Flip on a laptop for one-tap connect") rather than left to fail silently. WalletConnect integration is on the roadmap.

Roadmap (priority order): WalletConnect for mobile EVM; rate-limiting on /api/quote; server-side RPC proxy; quote signing for client-server integrity; inbound third- party security audit before sustained third-party traffic.

9. Conclusion

The cross-chain category has spent four years and approximately $2.5 billion in lost user funds proving that lock-and-mint architecture is structurally vulnerable. The category does not need a more secure custodian. It needs a different design.

Intent-based settlement provides that design. Solvers post collateral and compete to fill orders; failures are slashed; users are refunded without principal loss. The locked treasury — the dominant attack surface in the bridge category — does not exist. Flip applies this design to native Bitcoin and five additional production chains, delivered through a consumer interface that exposes the full cost structure of every quote and makes every settlement claim independently verifiable on public block explorers.

The protocol works. It is in production. Two reference swaps cited in this paper executed against the live system with on-chain receipts. The economic, security, and verifiability arguments above can be tested by any reader in under fifteen minutes with no special access.

Cross-chain swap is not a market that needs a better trust layer. It is a market that needs a refusal to require trust.

References

1. Chainalysis. Crypto Crime Report — Cross-Chain Bridge Attacks. (2024).
2. Elliptic. Tracking the Wormhole Hacker. (2022). Wormhole incident postmortem ($325M).
3. Ronin Network. Ronin Bridge Incident Disclosure. (2022). $625M loss; signature collusion mechanism.
4. Nomad. Nomad Bridge Postmortem. (2022). $190M replay vulnerability.
5. NEAR Foundation. NEAR Intents Protocol Specification, v1. (2025). Intent format, solver collateral, slashing condition.
6. NEAR Intents Documentation. 1Click API Reference. (2025). Quote, appFees, refund mechanism.
7. CoW DAO. Coincidence of Wants: Batch-Auction Settlement. (2023).
8. Uniswap Labs. UniswapX Whitepaper. (2023).
9. AUSTRAC. Guidance — Digital Currency Exchange Providers. (2024).
10. Chainalysis. Free Sanctions Screening API Documentation. (2025).

Appendix A: On-chain verification receipts

The two reference swaps cited in Section 6 can be verified via the following on-chain transactions:

• Swap 1 (Eth → Base ETH) deposit: 0xbf83615695ec2003485df602050ee59053fe2daee51f56bde24e7f9b2aff319b
• Flip operating fee account on NEAR: largevinegar3245.near

This document was prepared by Flip Pty Ltd. Published May 2026. Distribution unrestricted. Comments and corrections may be sent through the contact form. The canonical Markdown source is WHITEPAPER.md at the repository root.