Open Source - v0.1.0

Governance that sits between
AI and action

SignalWeaver evaluates AI decisions against declared policy, produces a deterministic outcome, and records a replayable trace for audit.

proceedgaterefuse

See It In Action

Interactive preview based on the live SignalWeaver demo flow.

SignalWeaver Demo

Interactive preview based on the live SignalWeaver demo flow.

Deterministic

Same input, same output. Always.

Model-Agnostic

Works with any AI system. No retraining required.

Full Traceability

Every decision auditable

Counterfactual Replay

Simulate policy changes

The Problem in One Sentence

AI agents make decisions that cannot be audited, traced, or governed - and regulators are now demanding answers that model-based systems cannot provide.

No Audit Trail

Neural networks are black boxes. When something goes wrong, there is no way to reconstruct why.

Compliance Gap

EU AI Act requires explainability and human oversight. Most AI systems cannot provide either.

No Policy Simulation

Cannot test "what if we changed the rules" without retraining entire models.

Core Primitives

Five building blocks that work together to govern AI agent behavior with mathematical certainty.

Truth Anchors

Policy rules that cannot drift or degrade over time. Deterministic by design.

  • Immutable policy definitions
  • Version-controlled rules
  • No model training required

Gate Engine

Real-time action evaluation. Every agent request passes through policy gates.

  • Allow, deny, or modify actions
  • Sub-millisecond evaluation
  • Context-aware decisions

Decision Trace

Complete audit trail of every decision, with inputs, rules applied, and outcomes.

  • Tamper-proof logs
  • Queryable history
  • Export for auditors

Replay

Re-execute past decisions to verify behavior or reproduce issues.

  • Exact state reconstruction
  • Debug production issues
  • Verify compliance retroactively

Counterfactual Replay

Simulate how policy changes would have affected historical decisions — impossible with model-based approaches.

  • Test policy changes before deployment
  • Measure impact on past decisions
  • A/B test governance strategies
  • Prove compliance to regulators

How It Works

Three steps from chaos to compliance

1

Define Policies

Write Truth Anchors in declarative YAML. Version control with Git.

2

Connect Agents

Route AI agent actions through the Gate Engine API.

3

Audit and Simulate

Query decision traces. Run counterfactuals. Prove compliance.

Get Early Access

SignalWeaver is currently in private beta. Request access to evaluate it for your AI governance needs.