1. What Axiom Is
Axiom Sovereign Engine is a cryptographic compliance evidence platform. It collects evidence from infrastructure and AI systems, evaluates that evidence against regulatory controls, and mints tamper-evident witness anchors that prove what happened, when, and under which policy. Every anchor is independently verifiable using standard SHA-256 -- no account, no vendor dependency.
The one-sentence version:
Axiom generates the compliance evidence that auditors, assessors, and regulators need to verify your systems -- without ever seeing your data.
The platform operates on a hash-only data model. Raw prompts, model weights, PII, and sensitive IP never reach the platform. Only cryptographic hashes and numeric factors are stored. This means near-zero liability for both the platform operator and the customer.
Axiom has two layers. Understanding the boundary helps you choose the right integration approach for your environment.
SWT3 (Protocol)
The Open Standard
Published to 10 registries. Free, open, language-agnostic. Works with or without the platform.
- Witness anchor format and fingerprint formula
- SDKs (Python, TypeScript, Rust, C#, Ruby, Swift, Go)
- MCP server (59 tools for AI agents)
- Clearing levels (data minimization)
- 266 procedures across 75 namespaces
- Public verification (client-side SHA-256)
- CLI:
swt3 init, swt3 procedures, swt3 demo
The analogy:
SWT3 is to Axiom what Git is to GitHub. The protocol is open and runs anywhere. The platform adds collaboration, dashboards, automation, and enterprise features on top.
3. How It Works
Four steps, from evidence to verification:
1
Collect
The Axiom CLI scans your infrastructure. The SWT3 SDK records your AI system behavior. Both produce numeric evidence factors.
2
Evaluate
The platform evaluates evidence against deterministic rules. Same inputs always produce the same PASS or FAIL verdict. No human judgment in the loop.
3
Anchor
Each verdict mints a tamper-evident SWT3 witness anchor with a SHA-256 fingerprint. The anchor is written to the compliance ledger and rolled into daily Merkle roots.
4
Verify
Anyone can verify an anchor independently using standard SHA-256. No account, no API key, no trust in the platform required. Auditors verify your evidence directly.
4. Core Services
These are the services the Axiom platform provides to subscribers, organized by what the platform does.
Evidence Collection
Passive evidence gathering from infrastructure and AI systems. Axiom reads configuration state and system output -- it never probes networks, opens ports, or injects traffic.
- Configuration state collection: 123 infrastructure controls evaluated by reading system configuration files, service status, and package manifests on an automated cycle (CA-7 continuous monitoring)
- Manual attestation: 42 policy and procedural controls attested by authorized personnel with non-repudiation
- External artifact ingestion: CKL, CKLB, Nessus, and PDF artifacts anchored via amnesia protocol (hash-only, original content never stored)
- AI inference witnessing: SDK-based recording of model calls, tool use, guardrail events, and drift via SWT3 anchors
Evidence Evaluation
Every piece of evidence is evaluated against a published rule, producing a binary PASS/FAIL verdict with a cryptographic witness anchor.
- Rule engine: Each control has a deterministic evaluation rule -- same inputs always produce the same verdict
- SWT3 anchor minting: Each verdict produces a tamper-evident anchor with SHA-256 fingerprint
- Automated alerting: FAIL verdicts trigger immediate notifications to your team
- Sovereign Score: Real-time compliance posture calculated from latest verdicts
AI Witnessing
Independent, cryptographic recording of AI system behavior for regulatory compliance.
- 266 procedures across 75 namespaces (inference, drift, guardrails, tools, identity, access, revocation, fairness, and more)
- Clearing levels: 4 tiers of data minimization (Level 0: analytics through Level 3: classified)
- Model drift detection: Rolling 24-hour pass rate monitoring with configurable thresholds
- Multi-provider support: OpenAI, Anthropic, AWS Bedrock, Ollama, Vercel AI SDK, LiteLLM (100+ models)
- MCP server: 59 tools for AI agent compliance via Model Context Protocol
Compliance Artifacts
Export-ready documents for auditors, assessors, and authorizing officials.
- OSCAL exports: NIST-validated SSP, POA&M, and Assessment Report (JSON, machine-readable)
- Executive summary: Boardroom-ready HTML report with Sovereign Score, KPIs, and framework coverage
- Gap-to-Green analysis: Evidence-based compliance gap identification with current-state findings, coverage trends, and 30-day posture history
- Auditor bundles: Policy + traceability matrix + regulatory appendix (stitched PDF-ready)
- W3C Verifiable Credentials: Ed25519-signed compliance passports (Enclave+ tier)
- C3PAO templates: Traditional HTML SSP and CSV POA&M for assessors who prefer non-OSCAL formats
Continuous Monitoring
Automated compliance health tracking that runs without human intervention.
- Drift detection (CA-7): Compares scans against baseline, classifies transitions (CRITICAL_DRIFT, REMEDIATION, STABLE, NEW)
- Vulnerability evidence aggregation (RA-5): Daily passive dependency analysis cross-referenced against known vulnerability databases, with findings recorded as POA&M entries. Milestone defaults (CRITICAL=7d, HIGH=30d, MEDIUM=90d) are configurable by the AO.
- POA&M lifecycle: Auto-open on FAIL, auto-close on fix, overdue alerting to your team
- Attestation lapse monitor: Automated review interval tracking with early warning before controls expire
- Merkle rollups: Daily cryptographic root of all anchors with RFC 3161 timestamps
Public Verification
Anyone can verify an SWT3 anchor independently. No account, no API key, no vendor dependency.
- Browser verification: Client-side SHA-256 at
/verify -- runs entirely in the browser
- Merkle proofs: Inclusion proof that an anchor existed in a specific daily rollup
- Passport verification: W3C Verifiable Credential validation with revocation status check
- Offline verification:
axiom verify --token SWT3-... works without network connectivity
Process Provenance
The cryptographic trust chain extends beyond anchor verification to the evaluation process itself. Auditors can independently verify that the rules, binaries, and pipeline that produced a verdict were unmodified.
- Rule determinism: Every verdict rule is a deterministic expression -- same evidence inputs always produce the same PASS/FAIL output. Rules are versioned in the verdict database with immutable evaluation logic.
- Binary integrity: The dependency analysis binary (Trivy) is pinned to a specific release version with a SHA-256 checksum verified before every execution. Never installed via rolling package managers or piped install scripts.
- Pipeline validation gate: Every deployment passes a 4-phase validation gate: secret scrub, evidence freshness verification, logic tests (87+ checks), and sovereign attestation. The gate itself is version-controlled and auditable.
- OSCAL validation: All compliance exports are validated against the NIST OSCAL reference implementation before anchor minting. An invalid export cannot produce an anchor.
- Auditor verification: A C3PAO can request the rule version hash, binary checksum, and gate execution log to independently confirm that the evaluation pipeline matches the documented process.
Framework Coverage
Cross-framework compliance from a single evidence set.
- 36 frameworks supported: NIST 800-53, CMMC, FedRAMP, NIST AI RMF, EU AI Act, SR 11-7, ISO 42001, HIPAA, OSHA-1926, and more
- Crosswalk engine: Map controls between frameworks automatically (e.g., NIST to CMMC)
- Framework-aware personas: Dashboard vocabulary, exports, and priority engine adapt to your primary framework
- UCT registry: 240 Unified Control Taxonomy procedures (public, namespace-aware)
5. CLI Quick Reference
Two CLIs exist. Use axiom for platform operations (scanning, reporting, evidence ingestion). Use swt3 for protocol operations (AI governance setup, procedure catalog, demos).
Rule of thumb:
If you are scanning infrastructure or generating compliance artifacts, use axiom.
If you are setting up AI witnessing, exploring procedures, or running demos, use swt3.
Axiom CLI -- Platform Operations (click to expand)
Installed on the server or customer enclave. Communicates with the Axiom platform securely.
| I want to... | Command |
|---|
| Scan all infrastructure controls | axiom scan --all |
| Scan a single control | axiom scan --control SC-7.6 |
| Scan without sending to platform | axiom scan --dry-run --all |
| Scan in air-gapped mode (local storage) | axiom scan --local |
| Check for drift since last baseline | axiom drift |
| Generate an offline HTML posture report | axiom report --local --html |
| Verify a single SWT3 anchor | axiom verify --token SWT3-E-... |
| Verify all enclave anchors | axiom verify --enclave |
| Adjudicate a control offline | axiom judge --procedure SC-7.5 |
| Ingest external evidence (CKL, Nessus, PDF) | axiom ingest --file evidence.ckl |
| Push verdicts to STIG Manager | axiom push --target stigman |
| Sync CISA KEV vulnerability feed | axiom kev --sync |
| Generate air-gap transfer bundle | axiom pulse --generate |
| Generate tenant provisioning SQL | axiom init --tenant-id ACME_PROD |
SWT3 CLI -- Protocol Operations (click to expand)
Installed via npm (npm i -g @tenova/swt3-ai). Works standalone, no platform account required.
| I want to... | Command |
|---|
| Set up AI governance for my project | swt3 init |
| Set up governance non-interactively (CI/CD) | swt3 init --profile eu-ai-act-high-risk |
| List all 266 procedures | swt3 procedures |
| Filter procedures by namespace | swt3 procedures --namespace MDL |
| Run a local demo (no API key needed) | swt3 demo |
| Check my governance config health | swt3 doctor |
| Check my governance config in CI (strict) | swt3 doctor --ci |
| Verify an anchor offline | swt3 verify |
| Generate a forensic audit timeline | swt3 audit |
| Check cloud compliance posture | swt3 status |
| See visual procedure coverage | swt3 status --coverage |
| View compliance changes over time | swt3 diff --since 7d |
| Set up a CI/CD compliance gate | swt3 gate --init --framework nist-ai-rmf |
| Generate a quick-start code snippet | swt3 quickstart |
6. Subscription Tiers
Open
Free
- 90-day evidence retention
- Infrastructure scanning
- AI witness anchors
- Public verification
- Standard API access
Pro
Monthly / Annual
- Unlimited retention
- All 123+ controls
- Auditor Share Links
- HTML + HMAC exports
- Compliance Passport
- 50 req/sec API limit
Enclave
Monthly / Annual
- All Pro features
- W3C Verifiable Credentials
- Regulatory Webhooks
- Merkle rollup access
- 99.5% SLA
- 200 req/sec API limit
Sovereign
ATO Sprint
- All Enclave features
- Air-gap deployment
- 1 business day SLA
- STIG Manager auto-push
- Custom SOW terms
- 500 req/sec API limit
Pricing details are available at /pricing. Sovereign engagements are sales-led with custom Statements of Work.
7. What Axiom Is NOT
Clear boundaries prevent misalignment in sales conversations and customer expectations.
Not a Security Scanner
Axiom witnesses compliance state. It does not discover vulnerabilities, perform penetration testing, or actively probe networks. Dependency analysis for RA-5 evidence reads package manifests passively -- it never opens ports, sends packets, or introduces attack surface.
Not an Approval Workflow
Axiom generates evidence. It does not approve systems, grant ATOs, or make authorization decisions. Authorizing officials and assessors use Axiom's evidence to make their own decisions.
Not a Model Evaluator
Axiom records what AI systems did, not whether they did it well. It does not benchmark model quality, accuracy, or performance. It witnesses the fact that an inference occurred and under what governance context.
The positioning:
Axiom is an independent witness. It proves compliance happened. It does not perform compliance, evaluate quality, or make decisions. The evidence it produces is cryptographically verifiable by anyone, including parties who do not trust the platform itself.