Three ways to get the outcome.

You're configuring a reference Customer Agent for this demo. Production agents submit the same request directly through the Margin402 API — no human required.

Job typeCode RepairDuration Parser Repairfunction parseDuration(input: string): number8-test verified outcome

Customer agentdemo-agent-····ActivePolicy: outcome-required · payment: x402

Each plan is a real execution policy the engine will follow — a different tradeoff between cost, confidence, and speed. The outcome is identical in all three: all tests pass, or you are refunded.

Lowest Cost

Variable risk

Minimize spend

Fixed price

$1.00

Strategy: Cheapest-first
Entry provider: Draft
First-attempt estimate: 35%
Escalation: After tier exhaustion

Cost priority

Reliability

Lowest upfront price
More attempts before success is likely
Slower expected completion

Execution policy: Strict cheapest-first ladder: Draft, then Repair, then Premium. Only one tier is active at a time — the engine may never skip ahead to a pricier provider while a cheaper tier has attempts remaining.
Retry: A failed tier consumes one of its attempts; the ladder advances only when the current tier has no attempts left.
Escalation: Escalation is exhaustion-driven, never economics-driven: the next tier unlocks only after the previous one is fully spent.
Verification: Every attempt runs the full 8-test suite — 6 visible, 2 hidden from providers. Binary: all pass or nothing ships.
If it can't finish: If the ladder is exhausted or the remaining budget cannot honourably continue, the contract is refunded.
Path: Draft → Repair → Premium

Recommended

Best Value

Balanced risk

Balance cost and confidence

Fixed price

$1.20

Strategy: Adaptive economics
Entry provider: Draft
Decision model: Re-ranked every round
Goal: Lowest expected cost to success

Cost efficiency

Adaptivity

Best expected cost-to-success
Adapts to real-time results
Can reject affordable-but-inferior options

Execution policy: All strategies with attempts remaining are available every round. The engine selects the lowest expected cost-to-success option — and will decline an affordable provider when a better expectation exists.
Retry: Re-evaluated per attempt using live results; failing tests feed back into repair-style strategies (visible failures only).
Escalation: Purely economic: escalate when a pricier provider becomes the cheapest expected path to a verified outcome — even mid-run, even above its list price.
Verification: Every attempt runs the full 8-test suite — 6 visible, 2 hidden from providers. Binary: all pass or nothing ships.
If it can't finish: The honouring rule applies: paying anyway beats refunding whenever the loss from delivering is smaller than the loss from stopping.
Path: Engine-decided each round

Highest Confidence

Low risk

Maximize certainty

Fixed price

$1.35

Strategy: Reliability-first
Entry provider: Premium
First-attempt estimate: 85%
Goal: Fastest path to verification

Reliability

Speed

Highest first-attempt pass rate
Fewer attempts on average
Higher upfront price

Execution policy: Reliability-first ladder: Premium, then Repair, then Draft. The highest-success provider is always attempted before any cheaper fallback is considered.
Retry: Premium holds up to three attempts at its current market price before any fallback tier unlocks.
Escalation: Fallback preserves reliability ordering — mid-tier Repair is preferred over the low-cost Draft whenever both are available.
Verification: Every attempt runs the full 8-test suite — 6 visible, 2 hidden from providers. Binary: all pass or nothing ships.
If it can't finish: The honouring rule applies: when the only path to done costs more than the budget that remains, Margin402 pays anyway while losing less than a refund would cost.
Path: Premium → Repair → Draft

Provider market

simulated provider market — every payment against it settles for real

Draft: 35% success estimate; Low cost
Repair: 45% success estimate; Feedback-aware
Premium: 85% success estimate; Highest reliability

Not sure? The engine recommends Best Value — minimizes expected cost-to-success while keeping every option available — the strongest balance of price, probability, and adaptivity.

Provider pricing follows a simulated provider market; every payment against it settles for real on Algorand Testnet.