Stablecoin escrow used to mean wiring USDC to a custodial intermediary and trusting them to release on cue. In 2026, conditional settlement is split across at least six distinct patterns: timelocks, multi-sig conditions, oracle-triggered release, milestone tranches, atomic swaps, and HTLCs. Each pattern has a canonical implementation (Sablier, Safe, Circle Programmable Wallets, Liquality, and a handful of others), and the right pick depends on counterparty trust, jurisdiction, and the asset on the other side of the trade.
​

This article walks through the six conditional-transaction primitives, compares the production implementations behind each, and maps them to B2B settlement use cases. Education-first: no single provider wins every scenario.
​

A conditional stablecoin transaction is a transfer where USDC, USDT, PYUSD, or another stablecoin is held in a programmable construct (smart contract, MPC vault, or custodial ledger entry) and released only when a defined condition is met. The condition can be time-based, signature-based, oracle-based, or event-based. Funds are not in the recipient's control until the condition resolves.
​

The construct matters because it shifts custody risk. A multi-sig Safe with a 2-of-3 condition is non-custodial. A Bridge.xyz escrow API is custodial. A Sablier stream sits in between: funds are locked in an audited contract, but the stream parameters cannot be modified after creation without the sender's key.
​

Time-locked stablecoin releases are the simplest conditional pattern. Funds vest linearly or cliff-release after X seconds. Sablier V2 supports both linear and dynamic vesting curves on Ethereum, Arbitrum, Optimism, Base, Polygon, Avalanche, BNB Chain, Scroll, Linea, and Sei. As of May 2026 Sablier reports over $5.1B in total streamed volume across more than 13,000 active streams.
​

Superfluid takes the same primitive and inverts it: instead of locking a balance and dripping it out, Superfluid wraps USDC into a "super token" (USDCx) that streams continuously by the second. Treasury teams at Gitcoin, Optimism Foundation, and Aave Grants use Superfluid for payroll because there is no per-period gas cost once the stream is open. The tradeoff is that Superfluid only works between accounts that both hold the wrapped Super Token, which adds an onboarding step for recipients.
​

Multi-sig escrow uses an M-of-N signature condition instead of a time condition. Safe (formerly Gnosis Safe) is the dominant implementation, with more than $100B in assets across deployed safes as of Q1 2026. A typical 2-of-3 buyer/seller/arbiter setup lets the buyer and seller jointly release funds, and only invokes the arbiter if they disagree.
​

The key difference from Sablier: a Sablier lock releases automatically when the timer expires, even if the counterparty has not delivered. A Safe escrow does not release until signatures arrive. This makes multi-sig the right pattern for goods or services with no fixed delivery date, and timelocks the right pattern for vesting, refunds, or scheduled payouts where time alone is the release trigger.
​

Oracle-triggered escrow releases funds when an external data feed reports a specific event. Chainlink Functions, Chainlink Price Feeds, UMA's Optimistic Oracle, and Pyth are the four production-grade options. Use cases:

UMA's Optimistic Oracle is interesting because it is not a feed. It accepts a human-readable assertion ("did Vendor X deliver invoice 47?") and lets any participant dispute the answer within a challenge window. Across uses UMA for cross-chain bridge attestations; Polymarket uses it for prediction-market resolution.
​

Milestone escrow splits a single agreement into multiple conditional releases. The simplest implementations are bespoke Solidity contracts deployed per deal. Production tooling includes Hookless (an escrow protocol designed for freelance and milestone work), Request Network's invoicing layer, and Superfluid's "scheduled streams" feature.
​

Milestone payouts are the most common pattern for B2B service contracts: 20% on signing, 30% on draft delivery, 50% on final acceptance. Each tranche is its own conditional release, typically gated by a multi-sig signature from the buyer plus, optionally, an oracle attestation. The pattern composes: a Hookless escrow can hold all three tranches in a single contract, with each tranche pointing at a different release condition.
​

Hash time-locked contracts (HTLCs) are the original conditional-transaction primitive, originally designed for Bitcoin Lightning and now used for cross-chain swaps. The mechanic: party A locks USDC on Ethereum behind a hashlock H, party B locks BTC on Bitcoin behind the same hashlock H. The first party to reveal the preimage of H claims both sides; if the timer expires, both sides refund.
​

Liquality (recently relaunched as Liquality Wallet) is the most accessible HTLC implementation, supporting USDC, USDT, BTC, and ETH swaps. THORChain and Maya Protocol use a related pattern (continuous liquidity pools with timelock fallback) for native-asset swaps without wrapping. For pure stablecoin-to-stablecoin swaps, atomic swaps are mostly obsolete: Uniswap V4, Curve, and CCTP V2 deliver lower-friction outcomes. HTLCs survive where one side of the trade is BTC or another non-EVM asset.
​

Not every conditional transaction needs to be onchain. Bridge.xyz (acquired by Stripe in October 2024 for $1.1B) offers a custodial escrow API where funds are held in regulated accounts and released by API call. The condition logic runs in Bridge's backend, not in a smart contract. This is the right pick when one or both counterparties cannot hold a non-custodial wallet, or when the deal needs to settle in fiat on either end.
​

Circle Programmable Wallets with the Policy Engine sit between custodial and non-custodial. The wallets are MPC-based and developer-controlled, but Circle's policy engine can enforce conditions (amount thresholds, allowlists, time windows, multi-approver workflows) at execution time. For deeper detail on policy engines, see stablecoin policy engines and compliance at execution time.
​

For most B2B settlement, multi-sig (Safe) plus a milestone wrapper (Hookless or a custom contract) covers 80% of cases. Time-locks (Sablier) handle vesting and scheduled refunds. Oracle triggers (Chainlink, UMA) come in when delivery confirmation is automatable. Atomic swaps are niche, useful only when one side is BTC. Custodial escrow (Bridge.xyz) is the fallback for fiat legs.
​

A pragmatic stack: Circle Programmable Wallets or a Safe for the underlying account, Sablier for any time-based release, Hookless or a bespoke contract for milestones, and Chainlink/UMA for oracle conditions. For invoice billing patterns that look conditional but are actually recurring, see the best stablecoin subscription billing APIs in 2026. For deal flows where the destination address itself encodes conditions, see programmable stablecoin addresses and smart address patterns.
​

Provider features and chain coverage cross-checked against Sablier docs (docs.sablier.com), Superfluid docs (docs.superfluid.finance), Safe docs (docs.safe.global), Circle's Programmable Wallets documentation (developers.circle.com), Liquality (liquality.io), Bridge.xyz developer portal (bridge.xyz), Hookless protocol docs, UMA docs (docs.uma.xyz), and Chainlink (docs.chain.link). TVL and volume figures from DeFiLlama as of May 2026. Bridge.xyz acquisition figure from Stripe's October 2024 announcement.
​