nomiarch
IN DEVELOPMENTBuilding the first proof

Infrastructure & AI.
Under your authority.

Intelligence can propose. Your policy decides.

We’re building an open source trust layer that keeps your systems within approved boundaries, controls what AI can do, and produces evidence of every change.

Explore Nomiarch See the first milestone
INTELLIGENCE ≠ AUTHORITYProposed architecture
01 / OBSERVEYour environment

Read reality.
Detect the drift.

Read-only access
02 / PROPOSELocal intelligence

Explain the finding.
Propose a repair.

No execution authority
03 / AUTHORIZEYour policy.
Your approval.

Permit only the
exact approved action.

Independent of the model
04 / EXECUTEBounded action

Apply the change.
Stay within scope.

Scoped credentials
05 / VERIFY & PROVEVerifiable evidence

Check the result.
Preserve the record.

Independent verification
Designed to operate inside your environmentThe model never authorizes itself.
Infrastructure + AIPolicy before actionHuman authorityEvidence by design
01 / THE PLATFORM

A finding is only
the beginning.

Security tools identify problems. AI can suggest a fix. Nomiarch is being designed to connect the decision, the action, and the proof—so you can see why a change was allowed and whether it worked.

[ 01 ]

Declare the boundary.

Define where workloads may run, what data they can access, and which actions require approval. Make the operating rules explicit.

POLICY · IDENTITY · SOVEREIGNTY
[ 02 ]

Control the action.

Let AI investigate and propose. Use independent authorization and a scoped executor to govern what can actually change.

AI · AUTHORIZATION · EXECUTION
[ 03 ]

Prove the result.

Verify the deployed state. Connect observations, approvals, changes, and results in an evidence package you can inspect and retain.

VERIFICATION · EVIDENCE · RECOVERY
Inside the architecture
YOUR ENVIRONMENT. YOUR BOUNDARY.

Run where you’re
allowed to run.

The long-term design spans public cloud, sovereign platforms, and disconnected environments. Your infrastructure stays in place. Nomiarch provides the policy and evidence layer across it.

Explore the deployment plan
01

Cloud & private infrastructure

Begin with Azure and containerized services on a Linux VM.

FIRST TARGET
02

Restricted & sovereign

Extend policy to location, administration, keys, and egress.

PLANNED
03

Disconnected & air-gapped

Local models, controlled artifact import, and offline evidence.

PLANNED
02 / THE FIRST PROOF

One real fix.
The entire trust loop.

The first milestone is deliberately specific: detect an Azure Storage account that allows anonymous blob access, propose a repair, authorize it, and verify the result.

STORAGE ACCESS / PROPOSED CHANGEIllustrative example

// Azure Storage account configuration

allowBlobPublicAccess: true
allowBlobPublicAccess: false
PROPOSAL
Disable anonymous blob access
AUTHORITY
Policy + human approval
EXECUTION
Bound to the approved change
PROOF
Independent read-back + evidence

Design target · No runnable release yet

  1. Observe the drift.

    Compare actual configuration with declared policy and identify the source that owns it.

  2. Propose, then authorize.

    Generate a bounded repair. Bind independent policy and human approval to that exact change.

  3. Execute and verify.

    Apply through scoped access. Read the deployed state again to check that the condition changed.

  4. Preserve the evidence.

    Record the decision and result, with a verifiable local export for recovery and review.

03 / BUILDING IN THE OPEN

Progress you
can inspect.

Start with a clear design. Build a narrow proof. Make it reproducible. Publish what works and what still needs to be tested.

01NOW / DESIGN

Define the boundary.

Trust model, service boundaries, VM deployment, and acceptance criteria for the first proof.

Read the architecture
02NEXT / PROTOTYPE

Close the loop.

Connect observation, local AI, policy, approval, execution, and verification in a test environment.

03THEN / RELEASE

Make it reproducible.

A versioned bundle, tested installation guide, refusal tests, and a recovery walkthrough.

Beyond the first proof: more infrastructure adapters, AI tool admission, sovereignty profiles, and disconnected operation.

A FEW GOOD QUESTIONS

Understand
the intent.

What is Nomiarch?

An open source project building a trust and authorization layer for enterprise infrastructure and AI. The goal is to keep systems within their approved operating boundaries and generate evidence of the decisions and changes made.

Who is it being built for?

Platform engineers, security teams, and technology leaders responsible for sensitive systems. The initial focus is Canadian government and regulated enterprise use cases, with policy profiles designed to expand over time.

Can I run it today?

There is no runnable release yet. The architecture and deployment plan are available now. The first release is intended to package containerized services for a supported Linux VM, with tested installation and recovery instructions.

Does it certify compliance or sovereignty?

Nomiarch is being designed to enforce declared controls and produce evidence for review. Authorization and certification remain decisions for the responsible organization. Sovereignty also depends on hosting, jurisdiction, administration, keys, and dependencies.

Is it only for air-gapped AI?

No. The scope includes conventional infrastructure and AI across connected, restricted, and disconnected environments. Fully air-gapped operation is a planned deployment mode, with controlled artifact import and no external service dependency in the internal runtime.

NOMIARCH

Give intelligence
a clear boundary.

Explore the vision. Inspect the design. Follow the first proof.

Read the documentation What’s coming first