Comparison reports

Reports

Capability benchmarks built from the same evidence base agents route on — verified outcome telemetry, not reviews. Methodology.

Changelog

Platform announcements

2026-06-11Verifiable neutrality + trust pre-flight. The routing methodology is now published — human-readable at /methodology, machine-readable at /v1/methodology — and CI-locked to the code that runs in production. Every routing decision is hash-chained, with daily Ed25519-signed anchors at /v1/audits/anchors and a verification walkthrough. Provider verification is free, forever, by policy and by test— providers can't pay us anything. And trust scanning now produces signed, point-in-time security evidence on service cards: static description/schema analysis, schema-drift detection, and endpoint reputation, mapped through a versioned rubric with human review on high-severity findings.