AI output is not authority.

Phionyx is a deterministic governance runtime for AI-generated actions. It preserves how a decision was governed — policy, state, action, verdict, and evidence — as a signed, replayable record.

Phionyx does not make the model deterministic. It makes the governance path deterministic.

What makes it different

One inspectable record — not four separate tools.

Phionyx is a deterministic governance path for AI-generated actions. Before an agent's output becomes a real action, it passes four checks — and binds them into a single record you can replay and verify:

1 · Policy gates

Rules decide what is allowed to proceed.

2 · State transition

The change is applied in a defined, repeatable way.

3 · Calibrated abstention

When it can't back an answer, it refuses or defers instead of guessing.

4 · Replayable evidence

A signed record of what happened — re-checkable by anyone, offline.

Other tools capture part of this — a signed receipt, or an audit-and-replay log. Phionyx unifies all four in one artifact, runs on school-safe consumer apps, and governs the AI that builds Phionyx itself.

What each profile addresses

Three runtime profiles.

Each profile addresses a distinct governance challenge — not a monolith. Each does one job; together they keep AI output accountable to evidence.

The notary

Evidence Runtime

A signed, replayable record of every governed decision.

The boundary

Abstention & Boundary

Hedge, defer, or refuse — with a calibrated confidence, not a guess.

The gate

Safety Gate

Fail-closed gates between model output and action, with escalation to a human.

See the three runtime modes — how each works →

What changes

The difference is what happens before action.

Without Phionyx

Model output
Action

The model's suggestion can become the system's behaviour. No deterministic record of scope, safety, approval, or replay.

With Phionyx

Model output
↓ gates
State + audit
↓ approval
Bounded action

Every transition passes verifiable gates and produces a replayable audit envelope. The model stays probabilistic; the path is reproducible.

Real-world demonstrations

Applications

A model saying “fixed” is not evidence.

It is a claim.

Runtime evidence for agentic AI. Tool calls · agent claims · gates · audit chains · replayable runs.

See how the evidence verifies →
Phionyx — Deterministic Runtime Evidence for Agentic AI Systems