Fleet Simulator

The Fleet Simulator (F-22) runs a realistic multi-department agent fleet with real governance evaluations through the actual UCS engine. 1,000 governed agents. Real audit trail. Live dashboard integration.

Quick Start

nomotic simulate fleet --agents 1000 --watch

This provisions 1,000 agents across 10 departments, starts continuous governance evaluations, and opens the dashboard for live monitoring.

Subcommands

Command
Description

nomotic simulate fleet

Start a fleet simulation with live dashboard

nomotic simulate batch

Run a batch of evaluations without live display

nomotic simulate stop

Stop a running simulation

nomotic simulate status

Check simulation status

nomotic simulate report

Generate a simulation report

Departments

The simulator distributes agents across 10 departments, each with realistic action profiles:

Department
Archetypes
Example Actions

CRM

customer-experience, sales-agent

read customer_record, update crm_record, export contact_list

Finance

financial-analyst, fraud-detection

read portfolio_data, execute market_trade, transfer funds

DevOps

devops-agent, operations-coordinator

deploy service, restart container, delete production_volume

Data & Analytics

data-processor, document-processor

read analytics_db, query data_warehouse, delete raw_data

Security

security-monitor, fraud-detection

scan network_segment, analyze incident_log, block ip_address

Human Resources

hr-agent, executive-assistant

read employee_record, update compensation_data, read payroll_db

Procurement

procurement-agent, operations-coordinator

read vendor_catalog, issue purchase_order, execute bulk_order

Support

customer-experience, sales-agent

read ticket_data, issue refund, export support_history

Legal

legal-assistant, research-analyst

read case_files, draft contract, export confidential_docs

Executive

executive-assistant, financial-analyst

read board_materials, access strategic_plan, export investor_data

Each department includes intentionally risky actions (marked with higher-risk targets) to generate realistic deny/escalate verdicts.

Behavior Profiles

Agents are assigned one of four behavior profiles:

Profile
Weight
Initial Trust
Risky Action Multiplier

Compliant

65%

0.70–0.90

0.2x (rarely picks risky actions)

Borderline

20%

0.45–0.65

1.0x (normal distribution)

Drifting

10%

0.55–0.75

2.5x (gravitates toward risky actions)

Rogue

5%

0.30–0.50

4.0x (heavily biased toward risky actions)

Dashboard Integration

Use --sim-mode with the dashboard to wire simulation data into the UI:

The dashboard shows live simulation data in the Agent Roster, Evaluation Feed, and Drift Alerts panels.

API

Last updated