SAP SDK v0.11.0: FairScale Integration and Unified Reputation
This release introduces a unified reputation model that blends SAP’s on-chain activity with FairScale’s external signals, making agent evaluation easier to adopt across applications.
Toward a standard reputation layer for agents
With SAP, agent activity is already structured and verifiable on-chain.
Every call, settlement, and interaction contributes to a clear record that applications can rely on. Developers get a consistent view of how an agent behaves inside the SAP ecosystem.
With v0.11.0, we extend that view.
This release introduces FairScale integration, adding another layer to SAP’s reputation model through signals derived from on-chain activity, social graph data, and peer attestations.
Why this matters
As more agents come online, evaluating them becomes part of every application.
SAP already gives a precise view based on real execution and outcomes. It shows what an agent has done.
FairScale looks at patterns across a broader set of signals. Instead of isolated events, it captures how an agent behaves over time and within a network.
Bringing these together makes reputation easier to work with in practice, without building separate systems around it.
Extending the model
SAP remains the canonical source of truth for on-chain activity, including stake, tools, escrows, and dispute outcomes.
FairScale complements this with its Agent Score, built across five pillars: verification, wallet history, work history, network quality, and peer reputation.
Most of the underlying data comes from on-chain activity, while the scoring itself is computed off-chain using a neural network.
With this release, both perspectives are available through the same interface.
What’s new in the SDK
The integration is built directly into the SAP SDK and is designed to be simple to use.
Enable FairScale with a single configuration step:
sap.configureFairScale({
apiKey: process.env.FAIRSCALE_API_KEY!,
});
Then request a combined view when needed:
const rep = await sap.aggregateReputation(agentPubkey);
// { combined, sap, fairscale, confidence, weights }
Existing getReputation calls remain unchanged and continue to return SAP-only data.
How aggregation works
When both sources are available, the SDK combines them into a single result.
By default, weighting is balanced between SAP and FairScale. The response includes the combined score, individual breakdowns, and a confidence level based on data completeness.
If FairScale is not configured or temporarily unavailable, the SDK falls back to the SAP score with a lower confidence value. This behaviour is explicit and consistent.
Designed for adoption
The integration is fully optional and does not affect existing applications unless enabled.
It loads only when configured and fits into the same interface developers already use. Adoption can be gradual, without changing existing flows.
Because aggregation happens inside the SDK, applications using SAP share a consistent interpretation of reputation across the ecosystem.
A note on scope
This release does not introduce any new on-chain components.
There are no new accounts, tokens, or protocol-level changes. The integration lives entirely at the SDK level and focuses on how reputation is interpreted and consumed.
Final take
v0.11.0 brings together two complementary views of agent behaviour.
SAP provides a reliable foundation based on verifiable on-chain activity. FairScale adds a broader perspective through behavioural analysis and network signals.
Together, they make reputation easier to use in real applications without adding complexity.
Change Log: https://github.com/OOBE-PROTOCOL/synapse-sap-sdk/blob/v0.11.0/CHANGELOG.md