Nomotic Terms.

An AI system that takes actions such as querying databases, sending messages, processing data, or making decisions on behalf of a person or organization. Unlike a chatbot that responds to questions, an agent acts. Nomotic can handle both chatbot and action-oriented agents.

A cryptographically signed identity record issued to an agent when it registers with Nomotic. It establishes who the agent is, what it is authorized to do, and binds its identity to its governance configuration. Every governed agent has one.

A portable, cryptographically signed bundle that packages an agent's accumulated operational intelligence (behavioral fingerprint, trust trajectory, organizational context, and distilled insights), into a platform-agnostic format that travels with the agent. AIR is an open standard (CC BY 4.0) published at agentpk.io/specs/air, enabling any conformant platform to receive and rehydrate an agent's behavioral history without requiring the issuing platform.

The full Nomotic platform, organized around six pillars that cover everything from agent identity through lifecycle, governance, context, orchestration, and intelligence. The complete system for managing AI agents at any scale.

A standardized packaging format for AI agents, similar to how a container image packages an application. A .agent file bundles everything an agent needs to run — its code, dependencies, configuration, and metadata — into a single portable file. Built with agentpk, the .agent format makes it easy to register, deploy, and govern agents across any platform.

A governance verdict meaning the requested action passed evaluation across all 20 dimensions and is approved to execute. The most common verdict for well-configured agents operating within their defined scope.

A 0 to 1 measure of how different an agent's recent behavior is from its own established baseline. Scores above 0.3 warrant attention. Scores above 0.4 indicate a significant behavioral shift that should be investigated.

A pre-configured governance template that defines appropriate behavior for a specific type of agent, such as financial transactions, healthcare, customer support, or HR operations. Nomotic includes 25 built-in archetypes. Assigning the right archetype calibrates governance for your agent's role automatically.

An immutable, hash-chained record of every governance evaluation ever performed. Each entry is cryptographically sealed so any modification, deletion, or reordering is detectable. Your permanent proof that governance happened.

A configurable lookback window of 10, 25, 50, or 100 evaluations that gives Nomotic memory of how an agent has been behaving. Used to detect trends, anomalies, and patterns that a single evaluation would miss.

Nomotic's core concept. A layer of infrastructure that sits in the execution path of AI agents and governs what they should do, not just what they are technically permitted to do.

A measurable shift in an agent's behavior patterns over time, such as changes in action types, targets, timing, or governance outcomes that diverge from the agent's established baseline. Some drift is normal. Significant drift triggers alerts.

See Agent Certificate. A cryptographically signed identity record issued to an agent when it registers with Nomotic.

An organizational knowledge document injected into governance evaluations. Examples include brand guidelines, compliance rules, domain vocabulary, and behavioral boundaries. Context profiles ensure governance reflects your organization's specific requirements.

A governance verdict meaning the requested action failed evaluation and is blocked from executing. The agent is notified, the action does not proceed, and the decision is recorded in the audit trail.

One of 20 behavioral criteria evaluated for every agent action. Each dimension scores the action from a different perspective, such as scope compliance, stakeholder impact, or ethical alignment. The combined scores produce the UCS.

A notification generated when Nomotic detects a statistically significant shift in an agent's behavior. Alerts surface in the Alerts page and trigger email notifications to org owners and admins.

A governance verdict meaning the action requires human review before it can proceed. The agent pauses, the action is queued for approval or denial by an authorized team member. Used when confidence is insufficient to Allow or Deny automatically.

The process triggered by an Escalate verdict. A pending item appears in the approval queue for human review. The reviewer approves the action or denies it. All decisions are logged in the audit trail.

A single governance assessment of one agent action. Takes under 3ms, scores the action across all 20 dimensions, produces a verdict of Allow, Escalate, or Deny along with a UCS score, and is sealed in the audit trail.

An Intelligence recommendation generated by Nomotic's analysis engine. Findings identify patterns in agent behavior that warrant attention, such as declining UCS scores, idle agents, verdict drift, or cost outliers. Each finding has a severity level and a suggested action.

All agents registered within your organization, viewed as a group. Fleet-level views show aggregate health, coverage, and behavioral trends across every agent you manage.

The system of policies, evaluation, and oversight that determines whether an agent action should be allowed, escalated for review, or denied. Nomotic's governance operates at runtime, before each action executes.

The percentage of your active agents that have received at least one governance evaluation in the current period. Low coverage means agents are running without oversight.

The cryptographic structure of the audit trail. Each record contains the SHA-256 hash of the previous record, creating a chain where any tampering is mathematically detectable.

An agent whose code is uploaded to and executed by Nomotic's infrastructure. Nomotic handles the runtime environment, dependencies, and execution. You provide the code.

A pattern where human reviewers in an escalation workflow become less effective over time, such as approving requests without scrutiny or consistently approving long sequences without any challenge. Nomotic monitors for human drift alongside agent drift.

The first pillar of the Nomotic AMP. Every agent must be registered and issued a certificate before it can be governed. Identity establishes who the agent is before asking what it should do.

The sixth pillar of the Nomotic AMP. Covers behavioral analytics, drift detection, governance recommendations, and the Behavioral Context Window. Intelligence learns from your agents' history to improve governance over time.

Nomotic's ability to stop an agent mid-execution if behavioral conditions change during a long-running task. The agent reaches a safe checkpoint before halting, preventing partial state or data corruption.

A compact, signed credential used to verify identity in API requests. Nomotic uses JWTs to authenticate agents and operators making API calls, ensuring every request comes from a verified source with the correct permissions.

The practice of periodically replacing API keys or signing keys with new ones. Regular key rotation limits the damage from a compromised credential. Nomotic supports key rotation without downtime — create a new key, update your integration, then revoke the old one.

The second pillar of the Nomotic AMP. Managing an agent from provisioning through deployment, monitoring, evaluation, and retirement. Lifecycle tracks which zone an agent is in and manages transitions between zones.

An agent registered and governed within the Nomotic platform, with a certificate, lifecycle management, governance evaluation, and audit trail. The opposite of a shadow agent.

The Agent Management Platform for AI agents. Derived from the Greek nomos (law, rule, governance), Nomotic represents an intelligent governance layer that defines the rules, boundaries, and constraints under which agentic systems operate. Every agent decision evaluated, every action accountable, every outcome auditable.

The fifth pillar of the Nomotic AMP. Covers scheduling agents, configuring triggers and webhooks, managing multi-agent workflows, and controlling execution priority and queue management.

Your company or team's account within Nomotic. All agents, certificates, evaluations, policies, and users belong to an organization. Each organization's data is fully isolated from others.

A configuration that defines governance behavior for your agents, including dimension weights, escalation thresholds, veto rules, and approval routing. Policies can be applied per agent, per archetype, or organization-wide.

The highest-trust zone in an agent's lifecycle. Agents in production are actively serving real workloads. Promotion to production requires governance verification.

The approval queue where escalated actions wait for human review. When an agent receives an Escalate verdict, the action is held in the queue until an authorized team member approves or denies it. Queue items expire if not reviewed within the configured time window.

Nomotic's agent discovery tool. Scans your organization's infrastructure for AI agents operating without governance, unregistered, uncertified, and outside policy. Surfaces shadow AI before it becomes a liability.

See Finding. An Intelligence recommendation generated by Nomotic's analysis engine identifying patterns in agent behavior that warrant attention.

The entry-level zone for agents in development and initial testing. Governance is active but thresholds are more permissive. A safe environment to build and test before moving to staging or production.

An AI agent operating within your organization without registration, governance, or oversight. Shadow agents are a security and compliance risk. Radar is designed to find them.

Nomotic's governing philosophy. Traditional systems control what agents can do. The Should Layer governs what agents should do, a distinction that becomes critical as agents take on more consequential work.

The middle zone between sandbox and production. Agents in staging are tested against production-like conditions with full governance enforcement before promotion.

A 0 to 1 score that reflects an agent's behavioral track record over time. Updated after each evaluation based on verdict outcomes and behavioral consistency. Higher trust scores can unlock more permissive governance thresholds.

A single 0 to 1 score representing the overall confidence level of a governance evaluation, calculated as the weighted composite of all 20 dimension scores. Scores above 0.7 typically produce an Allow verdict. Scores below 0.4 typically produce Deny. The middle range triggers escalation or closer review.

The outcome of a governance evaluation: Allow, Escalate, or Deny. Every evaluation produces exactly one verdict, which is sealed in the audit trail.

A governance dimension with the authority to block an action regardless of how well it scores on other dimensions. A scope violation cannot be overridden by strong scores elsewhere.

Value of Information escalation. Rather than escalating on a fixed threshold, Nomotic escalates when the expected value of a human decision exceeds the cost of the delay. Escalations happen when they matter, not by default.

An HTTP callback that delivers real-time event notifications from Nomotic to an external URL. Use webhooks to trigger other systems when evaluations occur, agents change state, or drift alerts fire.

An agent that runs on your own infrastructure and connects to Nomotic's governance layer via the SDK or API. You control the execution environment. Nomotic governs the decisions.

The cryptographic certificate standard that underpins Nomotic's agent identity system. Agent certificates follow the X.509 format, providing a globally recognized structure for identity, validity periods, and cryptographic signatures.

A designated checkpoint during agent execution where the governance layer can pause or stop the agent without leaving partial state. Yield points allow Nomotic's interrupt authority to work safely — the agent reaches a clean stopping place rather than being cut off mid-transaction.

The stage of an agent's operational lifecycle: sandbox, staging, or production. Zones determine governance thresholds and access permissions. Agents move through zones as they mature from development to production deployment.