The Behavioral Control Plane™
for your agent fleet.
Governance isn't a policy document. It's a decision made before every action, at runtime, in sub-milliseconds. Nomotic's Behavioral Control Plane™ evaluates every agent action across 20 dimensions simultaneously — with interrupt authority, VOI-based escalation, and tamper-evident audit records.
Not a guardrail.
A control plane.
Guardrails are passive — they wait for something to go wrong. A control plane evaluates every action before it executes, maintains authority throughout, and can intervene mid-stream. The difference between observability and governance is the commit boundary: can the layer say no?
- ✓ 20 dimensions evaluated simultaneously in under 1ms
- ✓ Veto dimensions that override all other scores
- ✓ Decision-theoretic verdicts backed by cost models
- ✓ Interrupt authority throughout execution, not just before
- ✓ VOI-based escalation — humans engaged when the math says so
- ✓ Hash-chained, tamper-evident audit trail for every decision
Complete governance infrastructure.
20-Dimensional Evaluation
Scope, authority, resources, behavior, impact, stakeholders, incidents, isolation, timing, precedent, transparency, human oversight, ethics, and jurisdiction — evaluated simultaneously.
All TiersVeto Dimensions
Some dimensions don't negotiate. Scope violation? Stopped. Ethical breach? Stopped. Jurisdictional violation? Stopped. No weighting overrides a veto.
All TiersVOI-Based Escalation
Escalate to humans when the expected value of their input exceeds the cost of delay. Not always. Not never. When the math says it matters.
EnterpriseInterrupt Authority
Stop agents mid-action. Cooperative rollback to safe state. Organizations deploy more confidently when they know they can stop anything, mid-stream.
All TiersHash-Chained Audit
Every evaluation cryptographically linked to the previous one. Tamper-evident. EU AI Act Article 12 ready. Complete reasoning artifacts captured.
Team+Counterfactual Replay
Replay any agent's history against alternative governance configurations. Answer 'what would have happened if' for every audit question a regulator asks.
EnterpriseSimple to integrate.
Powerful at scale.
from nomotic import GovernanceRuntime, Verdict
runtime = GovernanceRuntime()
verdict = runtime.evaluate(action, context)
if verdict.verdict == Verdict.ALLOW:
# Proceed — governance authorized
handle = runtime.begin_execution(action, context)
# Interrupt authority active throughout
for step in workflow:
if handle.check_interrupt(): break
do_work(step)
runtime.complete_execution(action.id, context)
# Verdict: ALLOW · UCS: 0.831 · 0.62ms · sealed
Governance features
across tiers.
| Feature | Community | Team | Enterprise |
|---|---|---|---|
| 20-dimensional behavioral evaluation | ✓ | ✓ | ✓ |
| Sub-millisecond evaluation (avg 0.8ms) | ✓ | ✓ | ✓ |
| Veto dimensions (scope, isolation, jurisdiction, ethics) | ✓ | ✓ | ✓ |
| Three-tier decision cascade | ✓ | ✓ | ✓ |
| Interrupt authority mid-execution | ✓ | ✓ | ✓ |
| Hash-chained audit trail | ✓ | ✓ | ✓ |
| Constitutional rules engine | ✓ | ✓ | ✓ |
| VOI-based escalation to humans | — | ✓ | ✓ |
| Custom governance dimensions | — | — | ✓ |
| Counterfactual replay | — | — | ✓ |
| EU AI Act reporting | — | — | ✓ |
Govern every action.
Before it executes.
The open-source core includes full 20-dimensional evaluation. Enterprise adds counterfactual replay, custom dimensions, and EU AI Act reporting.