From zero to governed
in under 15 minutes.
Install Nomotic, register your first agent, apply your first governance policy, and generate your API key. Everything you need to put a behavioral control plane on your agents today.
pip install nomotic
Install Nomotic
A single pip install with zero required dependencies.
pip 21+
Upgrade with pip install --upgrade pip
Internet access
For initial install only. Runtime has zero external calls.
Verify the installation:
Behavioral Control Plane™ — ready
python -m venv .venv && source .venv/bin/activate before installing. This keeps your Nomotic install isolated from other projects.
Register Your First Agent
Every governed agent starts with a birth certificate — a cryptographic identity bound to a governance policy.
Connect Nomotic to your agent and give it an identity. Nomotic supports any agent framework — LangGraph, CrewAI, AutoGen, or custom Python.
# Initialize the Behavioral Control Plane™
nomotic = Nomotic.init(
api_key="nom_your_api_key_here", # from Step 4
environment="development"
)
# Register your agent with a birth certificate
agent = nomotic.register_agent(
AgentConfig(
name="my-first-agent",
archetype="general", # or healthcare, finance, etc.
description="My first governed agent",
owner="your-team",
)
)
print(f"Agent registered: {agent.certificate_id}")
Certificate: NOM-8C4F2A1E
Archetype: general
Trust score: 0.500 (baseline)
Status: ACTIVE
Available archetypes — pre-built governance configurations for common use cases:
| Archetype | Use Case | Key Constraints |
|---|---|---|
general | General purpose agents | Balanced defaults across all 14 dimensions |
healthcare | Clinical, patient data | HIPAA-aligned, strict data isolation |
finance | Trading, risk, compliance | Jurisdictional veto, audit requirements |
customer-service | Customer-facing agents | Stakeholder impact, tone enforcement |
data-pipeline | ETL, data processing | Scope compliance, resource limits |
code-execution | Code generation/exec | Isolation integrity, scope veto |
Apply a Governance Policy
Define what your agent is allowed to do — and what it absolutely cannot.
A governance policy defines the 14-dimensional evaluation parameters for your agent. Start with a simple policy and extend it as your needs grow.
policy = GovernancePolicy(
name="my-first-policy",
# Hard veto rules — absolute, no weighting overrides these
veto_rules=[
VetoRule.SCOPE_VIOLATION, # agent must stay in scope
VetoRule.ISOLATION_BREACH, # no cross-tenant data access
VetoRule.ETHICAL_VIOLATION, # ethical constraint is absolute
],
# Weighted dimensions — scored 0.0 to 1.0
dimensions={
Dimension.BEHAVIORAL_CONSISTENCY: 0.8, # threshold
Dimension.STAKEHOLDER_IMPACT: 0.7,
Dimension.HUMAN_OVERSIGHT: 0.6,
Dimension.TRANSPARENCY: 0.75,
},
# Minimum Unified Confidence Score to proceed
ucs_threshold=0.70,
# Escalate to human when Value of Information exceeds cost of delay
voi_escalation=True,
voi_threshold=0.85,
)
# Apply to your registered agent
agent.apply_policy(policy)
print(f"Policy applied: {policy.policy_id}")
Policy ID: POL-2F8A19C4
Veto rules: 3 active
Weighted dimensions: 4 configured
UCS threshold: 0.70
VOI escalation: enabled at 0.85
The 14 governance dimensions evaluated on every action:
| Dimension | Type | What it governs |
|---|---|---|
| Scope Compliance | Veto | Agent stays within defined operational boundaries |
| Isolation Integrity | Veto | No unauthorized cross-context data access |
| Ethical Compliance | Veto | Hard ethical constraint enforcement |
| Jurisdictional Compliance | Veto | Regulatory and geographic constraints |
| Behavioral Consistency | Weighted | Actions align with established behavior patterns |
| Stakeholder Impact | Weighted | Downstream effect on affected parties |
| Human Oversight | Weighted | Human review requirements for sensitive actions |
| Transparency | Weighted | Action explainability and auditability |
| Resource Authorization | Weighted | Budget and resource limits respected |
| Temporal Compliance | Weighted | Time-bounded and scheduled constraint enforcement |
| Precedent Alignment | Weighted | Consistency with prior approved actions |
| Incident Detection | Weighted | Real-time anomaly and incident identification |
| Authority Verification | Required | Agent identity and permission confirmation |
| Trust Score | Required | Cumulative behavioral trust history |
Generate Your API Key
Authenticate your runtime connection to the Nomotic Behavioral Control Plane™.
API keys authenticate your agents to the Nomotic runtime. Each key is scoped to your organization and can be restricted to specific agents, environments, or permission levels.
--name "development" \
--env development \
--scopes govern,audit,read
API Key — development
ACTIVERun Your First Evaluation
Govern a real action and see the Behavioral Control Plane™ return a verdict.
Submit an action to the governance runtime and receive a verdict in sub-millisecond time. The verdict includes a full 14-dimensional breakdown and a Unified Confidence Score.
import os
nomotic = Nomotic.init(api_key=os.getenv("NOMOTIC_API_KEY"))
agent = nomotic.agent("my-first-agent")
# Govern any action before it executes
verdict = agent.govern(
action="read_customer_record",
target="customer_db",
context={
"purpose": "generate_report",
"requester": "pipeline-job-42",
"data_sensitivity": "medium",
}
)
# Check the verdict before proceeding
if verdict.allowed:
print(f"Proceeding. UCS: {verdict.ucs}")
# execute your action here
else:
print(f"Denied: {verdict.reason}")
What the verdict looks like:
The Governance Brief.
Industry analysis, competitive intelligence, and technical deep dives
on governing AI agents. Biweekly.