Machine-readable data
The public evidence layer behind AgentenCode.
Selected structured datasets are published so product claims can be inspected beyond the rendered web pages.
Public datasets
Evidence coverage: 129 of 130 agents have field-level claims. Relay.app (discontinued in September 2026) currently has no claims in the public evidence dataset. Missing evidence means unknown, not false.
Public versus private layers
The public layer contains reviewed profile data, evidence claims, governance signals, relations and history. Internal monitoring, review queues and source-registry workflow data are not exposed as public facts.
Semantics for developers and AI systems
Missing canonical fields are unknown, not false. Scope is part of the meaning of a claim, and verification dates are explicit. Consumers should not flatten enterprise-only evidence into a universal product statement.
The JSON layer is intended for transparent reuse and inspection. The linked primary sources remain the authoritative reference when products change.
Versioning and reuse
Consumers should treat the JSON files as a versioned public reference layer rather than a complete mirror of internal research systems. The files are designed for inspection, search integration, reproducible comparisons and machine-assisted retrieval.
When a claim is reused, preserve its scope and verification date. A field documented for an enterprise plan, a specific region or one deployment surface must not be generalized to every user of the product.
Data model and interpretation
The public datasets are intentionally normalized so the same concepts can be reused across profiles. An agent record provides identity and summary information; evidence records bind sources to specific canonical fields; governance records expose normalized trust signals; history records add temporal context; and relation records describe explicit links between agents, platforms or system components.
This structure is designed to support both human inspection and machine retrieval. It also reduces the risk that a prose summary silently changes the meaning of a source-backed fact. Applications that consume the data should preserve the original scope, confidence and verification metadata.
Stable identifiers and update behavior
profile identifiers and canonical field names are intended to remain stable across editorial updates. When a provider changes a feature, AgentenCode updates the affected evidence or history record instead of silently redefining an existing field. This makes downstream comparisons easier to audit.