How It Works

The Strix Proof Chain

Governance without proof is a memo. Strix produces structured, verifiable evidence for every governed decision — AI actions included.

Canonical Proof Flow

When a caller attempts a governed mutation, the request flows through the embedded governance kernel. The kernel evaluates risk, checks policy, and — for high and critical capabilities — requires a payload-bound execution token. The token binds approval to the exact SHA-256 hash of the canonical payload, the calling actor, and the session.

Three execution paths, one proof shape

Whether the decision runs through a token-validated high-risk flow, the external Strix SDK, or a local policy fallback, the output is the same: an evidence record with a sealed hash, a decision vocabulary (allow or deny), and a proof receipt injected into the mutation response.

Public verification

Anyone can hit the public proof API with an evidence ID. The server re-computes the canonical hash from the stored fields and compares it to the sealed record. A mismatch proves tampering.

Structural guarantees

17 structural invariants run on every CI build. They enforce zero raw mutation bypasses, exact count of governed mutations, absence of native confirmation dialogs, and the decision vocabulary. A single regression fails the pipeline.

Evidence durability

Evidence is written synchronously to the local governance store on every governed call, then pushed to the central Strix Platform API via a durable outbox with status tracking. A daily integrity cron runs verification sweeps across the entire chain.

For auditors and integrators

The same verification endpoints serve machines and humans. Point your auditor tools at the public proof API with any evidence ID.

GET /api/proof/:evidenceId
POST /api/proof/verify { evidenceId }