SREL is a real, open standard you can implement right now — CC0 spec, a published grammar and profiles, with an Apache-2.0 reference validator and conformance suite in preparation. The hosted Shadow Simplex scoring API is a design preview; we're opening early access rather than pretending it's a finished product.
ABNF grammar, TOML wire format, URN-identified profiles. Implement and emit conformant documents today; the reference validator and conformance suite are in preparation.
Run failure-mode probes in your own pipeline; emit SREL/AI-1 output. No data leaves your environment.
Hosted composite scoring, C(κ) normalization, register vectors. In preview — request access; not yet generally available.
A SREL/AI-1 document is structured, machine-readable risk output: capability normalizer state, the meta-condition veto vector, 54 primitive sections, register weights, and a composite band. Downstream monitoring and remediation systems ingest it directly.
SDK packages, a Terraform provider, and stable hosted endpoints are on the roadmap, not shipped. We're not publishing install commands for packages that don't exist yet. If you want to integrate scoring, request early access and we'll bring you into the preview with realistic expectations about calibration status.