SeenRelay
MEASURED COST AVOIDANCE

Prove the savings on your workload before you enable reuse.

SeenRelay targets one economic problem across agents, services, CI, browser workers and selected IoT/edge reads: independent callers repeatedly paying money, latency or constrained capacity for the same exact read-only execution. Shadow mode keeps every authoritative call enabled while you measure whether a cheaper safe path exists.

Start locallyInstall with an agentSee measured evidenceEconomics Lab
Hosted protocol remains free during bootstrap$0Keep SeenRelay only where measured net economics remain positive
THE DECISION EQUATION

Reuse rate alone is not enough.

Let N be protected validations, C the marginal cost of a full authoritative validation, r the measured fraction that policy can safely reuse, and H the per-call overhead of the cheaper path. Count provider spend, latency and constrained capacity only when they matter to the workload.

baseline cost ≈ N × C

reuse-path cost ≈ N × H + N × (1 - r) × C

net avoided cost ≈ N × r × C - N × H

Use SeenRelay only when:
  measured net avoided cost > integration + operating cost
  AND outcome equivalence remains acceptable.
WHERE THE MATH CAN WIN

Protect expensive repeated validation, not cheap requests.

The best candidate is deterministic, read-only, recurrent and materially more expensive than the validation shortcut. Local, source-native and provider-native mechanisms get credit before SeenRelay.

Browser / portal validationHeadless sessions, proxy time, rendering or multi-step navigation that repeatedly establishes the same bounded state.
Metered extractionCommercial scraping, parsing or structured extraction where each repeated validation consumes credits or capacity.
Paid search / rate-limited APIsRepeated source-backed checks with a real per-call or opportunity cost.
Multi-step validationFetch → render → parse → model chains where one safe reusable result can prevent downstream work.
Outside the target: cheap one-off requests, mutations, low-repeat workloads, or any path already solved by an equivalent authoritative cache or source-native validator.
NO NETWORK EFFECT REQUIRED

The first savings can stay entirely inside one customer's fleet.

The current local-first order is exact in-flight reuse → caller-owned encrypted private L1 → source-native confirmation → optional shared CHECK → original validation. A customer does not need other SeenRelay users before local/private/source-native savings can exist.

The public relay is an optional evidence layer, not the economic foundation of the first deployment.

Local / in-flightCoalesce the same eligible work already happening at once.
Private L1Reuse caller-owned encrypted state across workers or restarts under explicit freshness policy.
Source-nativePrefer ETag, Last-Modified or a stronger authoritative version mechanism when available.
FallbackIf evidence is insufficient, perform the original validation normally.
YOUR COST INPUTS

Use your invoice, not a public benchmark.

The only deployment evidence that matters is your own measured workload. Enter the costs and constraints you actually pay for, then compare them with the measured cheaper path.

Provider spendUse your real marginal cost per call, credit, page, query or request. If the bill is fixed, count savings only when usage changes tier, overage or required capacity.
LatencyCount wall-clock time only when it affects throughput, SLOs, user experience or scarce execution capacity.
Native controlsCredit existing caches, validators and provider-native shortcuts first. SeenRelay must beat the best equivalent path you already have.
Operational overheadInclude integration, runtime and maintenance cost. A technically reusable result is not commercially useful if the layer costs more than it avoids.
The commercial decision is deliberately workload-specific. SeenRelay does not use a public benchmark to promise your savings.
SHADOW FIRST

Measure one validation path before touching policy.

The deterministic clients can sit around an existing validation without suppressing it. During the measurement window, the original call remains authoritative and the report records recurrence and prospective economics.

Only an explicit caller policy can later authorize bounded reuse.

JavaScript / TypeScript
const validatePrice = relay.protectValidation({
  fact,
  validate: ({ conditionalHeaders }) =>
    expensiveValidation(conditionalHeaders)
});

const value = await validatePrice(knownValue);
THE COMMERCIAL RULE

If the workload does not save more than the layer costs, do not deploy the layer.

The first useful outcome of SeenRelay can be a negative one: it prevents a team from adding complexity where native controls or sparse recurrence already make the economics bad.

Measure a workloadProduct details