TruthOps Overview

TruthOps enables autonomous agents to handle routine AI governance tasks — from monitoring and alerting through remediation and reporting. Agents work semi-autonomously (requiring approval) or fully autonomously (within policy bounds).

TruthOps Dashboard

What is TruthOps?

TruthOps is a collection of specialized agents that continuously monitor your AI systems and take corrective actions:

Monitoring Agents: Track LLM outputs for hallucinations and policy violations
Remediation Agents: Auto-correct issues or quarantine problematic outputs
Compliance Agents: Generate evidence and audit trails for regulatory requirements
Reporting Agents: Create executive summaries and risk dashboards

Agent Types

Monitoring Agents

Automatically scan AI outputs continuously:

Detect hallucinations (94%+ accuracy using NLI)
Flag policy violations
Identify data leaks or sensitive info leakage
Monitor for brand misrepresentation
Track performance degradation

Autonomy Levels:

Alert Only: Flag issues, wait for human review
Auto-Escalate: Alert supervisors immediately
Snapshot: Take screenshots/logs for review

Remediation Agents

Automatically fix detected issues with human control when needed:

Generate corrections for hallucinations
Rotate API keys on leak detection
Quarantine problematic outputs
Retrain models on corrected data
Update system prompts to prevent recurrence

Autonomy Levels:

Manual: Wait for approval before acting
Auto-Act: Act immediately, log for review
Emergency: Act immediately on critical issues

Compliance Agents

Automatically maintain regulatory evidence:

Collect control evidence from 16+ system connectors
Generate audit trails for all AI decisions
Create compliance reports (SOC 2, ISO 42001, GDPR, etc.)
Track policy exceptions and approvals
Manage incident response playbooks

Autonomy Levels:

Manual Approval: Require sign-off before submission
Auto-Submit: Submit automatically, notify leadership
Scheduled: Submit on regular cadence (daily/weekly/monthly)

Reporting Agents

Automatically generate insights and dashboards:

Executive governance dashboards
Risk scoring and trending
Policy violation reports
Spend analysis and cost optimization
Team performance metrics

Autonomy Levels:

On-Demand: Generate reports when requested
Scheduled: Generate daily/weekly/monthly automatically
Event-Driven: Trigger on threshold breaches

Key Capabilities

Sub-200ms Enforcement

All agent actions (monitoring, remediation, compliance) execute in sub-200ms:

Non-blocking to user interactions
Fail-open (system continues if agent fails)
Circuit breaker protection

Audit Trail

Every agent action is automatically logged for audit purposes:

Who triggered the action (agent, user, schedule)
What happened (check performed, correction made, evidence collected)
When it happened (precise timestamp)
Why it happened (which policy/rule)
Outcome (success/failure/manual review)

Policy-Driven

Agents follow organizational policies:

Define what agents can and cannot do
Set approval thresholds
Configure escalation paths
Manage exceptions and overrides

Multi-Channel Alerting

TruthOps automatically notifies stakeholders via:

Email
Slack / Teams
PagerDuty
Custom webhooks
In-app dashboard

Available Tiers

Feature	Professional	Business	Enterprise
Monitoring Agents	—	✓	✓
Remediation Agents	—	Manual only	Auto+Manual
Compliance Agents	—	✓	✓
Reporting Agents	—	✓	✓
Autonomy Levels	—	Limited	Full
Custom Agents	—	—	✓
SLA & Uptime	—	99.5%	99.9%

Quick Start

1. Enable Monitoring Agents

Navigate to TruthOps → Agents
Click Enable Monitoring Agent
Configure:
- Check frequency (1hr, 4hr, daily)
- Severity thresholds
- Alert channels
Click Activate

2. Configure Autonomy

For each agent type, set autonomy level:

Monitoring Agent
  ├─ Alert Only (flag issues, humans decide)
  └─ Auto-Escalate (escalate to manager)

Remediation Agent
  ├─ Manual (approve all actions)
  ├─ Auto-Act (act, notify afterwards)
  └─ Emergency (act immediately on critical)

Compliance Agent
  ├─ Manual (require sign-off)
  └─ Auto-Submit (auto-submit, notify leadership)

3. Set Policies

Define what agents can do via policies:

monitoring_agent:
  alert_on:
    - hallucinations
    - policy_violations
    - sensitive_info_leakage
  severity_threshold: high
  escalate_to:
    - compliance_manager
    - ciso

remediation_agent:
  actions_allowed:
    - generate_correction
    - quarantine_output
    - rotate_keys
  require_approval:
    - modify_system_prompt
    - retrain_model
  emergency_actions:
    - quarantine_output
    - isolate_system

4. Monitor & Review

Dashboard shows:

Agent activity (checks performed, actions taken)
Issues detected and resolved
Approval queue (if using manual autonomy)
Audit trail and compliance evidence

Common Workflows

Auto-Correction of Hallucinations

LLM generates response
  ↓
Monitoring Agent automatically detects hallucination (NLI: 5% entailment)
  ↓
Remediation Agent automatically generates correction
  ↓
Correction automatically deployed in <4 seconds
  ↓
Audit automatically logged for compliance

Compliance Evidence Collection

Hour 00:00: Compliance Agent starts automatically
  ↓
Collects evidence automatically from 16+ connectors (AWS, Okta, ServiceNow, etc.)
  ↓
Generates SOC 2 / ISO 42001 audit trail automatically
  ↓
Hour 00:45: Evidence aggregated and verified automatically
  ↓
Auto-submits to compliance system or notifies leadership

Weekly Governance Report

Every Monday 09:00 AM (automatically)
  ↓
Reporting Agent runs automatically
  ↓
Collects automatically:
  - Hallucinations detected/resolved
  - Policy violations
  - Compliance evidence status
  - Cost and usage trends
  ↓
Generates executive summary
  ↓
Sends to leadership automatically via email + Slack

Security & Trust

TruthOps agents are built with security-first principles:

API Key Isolation: Each agent has separate, scoped credentials
Audit Trail: Every action is logged and immutable
Circuit Breaker: Agents fail-open (never block user traffic)
Explicit Approval: Critical actions require human sign-off
Rate Limiting: Agents respect rate limits and backoff
Monitoring: Agent behavior is continuously monitored

Next Steps

Agents: Learn about specific agent types
Autonomy: Configure autonomy levels
Configuration: Set up monitoring policies
Monitoring: View agent activity and audit trails