Governance infrastructure that prevents unsafe AI and automated actions from reaching production.
AI can propose anything. Waveframe determines whether the proposed action is allowed to execute.
We stop bad AI actions in one function call.
Waveframe is an integrated governance and enforcement platform with distinct surfaces for governance teams, developers, operators, and auditors.
Existing policy, SOP, or authority source
↓
Ledger Workspace
↓
Compiled authority
↓
Waveframe Cloud
↓
Waveframe Guard
↓
CRI-CORE
↓
ALLOW / BLOCK
↓
Production system
↓
Guard Inspector
- Ledger Workspace turns organizational policy into governed, publishable authority.
- Waveframe Cloud distributes immutable authority and preserves durable execution evidence.
- Waveframe Guard gives developers a simple runtime enforcement boundary.
- CRI-CORE deterministically decides whether a proposed mutation is admissible.
- Guard Inspector explains decisions, authority, and evidence to operators and auditors.
Guard is the developer adoption entry point. Ledger Workspace is the governance and enterprise expansion surface.
pip install waveframe-guardfrom waveframe_guard import Guard
guard = Guard.local(
authority="finance-policy@1.0.0"
)Guard loads published authority, constructs the governed proposal, invokes deterministic enforcement, blocks inadmissible execution, and preserves decision evidence.
Runtime enforcement remains local-first so a Cloud outage does not silently remove the execution boundary.
- A governance team starts with an existing policy, procedure, delegation-of-authority document, or other authority source.
- Ledger Workspace guides the team through normalization, diagnostics, compilation, review, and publication.
- The published authority receives a stable identity, version, and cryptographic hash.
- Waveframe Cloud distributes that authority to approved environments.
- An application or agent proposes an action through Waveframe Guard.
- Guard constructs a canonical proposal and submits it to CRI-CORE.
- CRI-CORE returns the kernel-owned
commit_alloweddecision. - Guard either executes the action or prevents it from reaching the production mutation boundary.
- Decision evidence is preserved for replay, inspection, and audit.
The system is model-agnostic. Proposed actions may originate from AI agents, traditional automation, applications, or human-operated workflows.
| Repository | Purpose |
|---|---|
| Waveframe Guard | Developer-facing runtime enforcement SDK |
| CRI-CORE | Deterministic execution-boundary enforcement kernel |
| Contract Compiler | Human governance to deterministic compiled authority |
| Proposal Normalizer | Canonical proposal construction and binding |
| Governance Ledger | Authority preparation and publication infrastructure |
These repositories provide the local developer path, deterministic protocols, and inspectable enforcement foundation.
| Product | Role |
|---|---|
| Ledger Workspace | Collaborative governance authoring, review, lifecycle, and publication |
| Waveframe Cloud | Authority distribution, organizational controls, and durable evidence |
| Guard Inspector | Decision, authority, receipt, and replay inspection |
| Multi-tenant Operations | Identity, tenancy, access control, supportability, and enterprise administration |
The commercial product is the integrated Waveframe platform—not any single repository in isolation.
Waveframe's institutional governance and methodology remain separate from runtime product responsibilities.
| Repository | Authority |
|---|---|
| Aurora Research Initiative | Governance authority and institutional accountability |
| Neurotransparency Doctrine | Epistemic doctrine and human decision ownership |
| Neurotransparency Specification | Structural compliance requirements |
| Aurora Workflow Orchestration | Governed workflow methodology |
These sources define governance expectations. Runtime products operationalize and enforce compiled authority without redefining it.
- The enforcement decision occurs before the production mutation.
commit_allowedis owned by the enforcement kernel, not reconstructed by callers.- Guard never compiles natural-language policy during runtime enforcement.
- Published authority is bound by identity, version, and cryptographic hash.
- Blocked actions do not reach the governed execution function.
- Local enforcement continues during Cloud unavailability.
- Cloud preserves evidence but does not replace the local execution decision.
- Inspector explains recorded decisions but does not rewrite them.
An AI system proposes reallocating $2 million between cost centers.
Without an execution boundary, the proposal may reach the financial system before a governance failure is detected.
With Waveframe:
Proposal received
↓
Authority resolved and verified
↓
Required approval missing
↓
commit_allowed = False
↓
Execution blocked
↓
Evidence preserved for inspection
The action never reaches the financial mutation boundary.
Version compatibility and dependency requirements are maintained at waveframelabs.org/compatibility.html.
