adapterOS
AI for CUI and classified work — with the evidence an assessor needs.
adapterOS runs deterministic inference on contract packets, intelligence products, and technical data inside the boundary — no egress in contested or air-gapped networks — and emits a replayable, cryptographically receipted record of every run.
- CMMC 2.0 (Level 2/3)
- NIST SP 800-171 / 800-172
- DFARS 252.204-7012 / 7021
- NDAA §1513 covered AI
- Air-gapped / IL4-IL5
Built for CMMC 2.0 and the coming covered-AI controls. Patent pending.
The governance gap
Cloud AI trades control for convenience. For regulated teams that creates a gap: network egress, opaque runtimes, and non-deterministic behavior collide with data residency, air-gapped operation, and audit requirements — so adoption stalls even when a model is approved in principle. adapterOS closes that gap by running in your boundary and producing evidence by default.
Unverifiable runtime
Unclear runtime behavior makes audit evidence hard to trust.
No change control
Ad-hoc deployment workflows make review and change control difficult.
Cloud-bound
Cloud-dependent runtimes are incompatible with air-gapped and classified environments.
What it helps teams do
Read and compare documents
Review contracts, reports, tickets, policies, spreadsheets, and technical records inside one approved workspace.
Ask within a focus
Configure a focus around a workflow so the system knows which dataset, policy boundary, and output shape are allowed.
Keep sources attached
Answers include citations, proof tier, and reviewer context so a person can inspect why the response says what it says.
Work locally
Pilot deployments are designed for local hardware, no routine sensitive-data egress, and explicit update handling.
What is real today
Working now
- ✓Source-backed answers with a proof tier on every response: exact, evidence, lineage, or degraded.
- ✓Deterministic replay for any past answer.
- ✓Authoritative token accounting. Estimates are not presented as billable usage.
- ✓Evidence-write durability. If sources are not persisted, the answer downgrades visibly.
- ✓Pilot validation. The latest strict-mode run passed all steps on May 3, 2026.
Pilot evidence
The credible commercial wedge is one workflow, one environment, one deployment scope, and one review path. The proof tier is the moat: adapterOS shows when an answer is exact, evidence-backed, lineage-backed, or degraded.
How it deploys
Hardware included
MLNavigator ships hardware pre-loaded with adapterOS. No separate procurement. No BYO hardware complexity.
Managed install
We install adapterOS in your environment, configure the source set and review path, and validate the workflow against your acceptance criteria.
Operator handoff
Your team operates adapterOS. Ongoing support includes update distribution, review-record retention guidance, and incident response.
Interfaces, inputs, and supply chain
Access
Use adapterOS through a web UI, a REST API, or the command line. Access is gated by JWT or API keys, with RBAC and policy packs defining boundaries before activation.
Inputs
Ingests PDFs, Markdown, and JSONL into an approved source set. No external model training; no data leaves the premises.
Supply chain
Every deployment ships SBOM artifacts for its components and signed run receipts for inference, so security teams can attest to what is installed and what ran.
What adapterOS does not do
- ✕Does not guarantee output correctness. Sources show why an answer was produced. Humans decide if the output is right.
- ✕Does not guarantee model safety or alignment. We govern how document work is scoped, cited, and reviewed. Model quality is upstream.
- ✕Does not make outbound network calls, send telemetry, or auto-update. Updates are explicit and verified.
- ✕Not a cloud document upload service, model marketplace, or open-ended consulting engagement.
- ✕Does not replace reviewer sign-off. adapterOS records evidence for review. It does not make the decision for the team.
Who adapterOS is for
Teams in controlled, disconnected, or export-sensitive environments — where reviewability, change discipline, and deployment control matter as much as model capability.
Defense & aerospace
ITAR-sensitive and CUI document work, CMMC-aligned evidence, air-gapped and classified deployment.
Critical infrastructure
Offline energy, transport, and utility operations that cannot depend on public cloud inference.
Healthcare & life sciences
PHI-bound document work kept on-premises, HIPAA-aligned, with an audit trail for every answer.
Regulated finance
Model-risk-governed AI with books-and-records capture and data residency inside the perimeter.
Government & public sector
Sovereign data and classified networks requiring local control and defensible operating records.
Evaluate adapterOS
Fixed-scope pilot. One document workflow. Hardware included. Deployment starts with the selected document, dataset, adapter, focus, and acceptance criteria.