GOVERNANCE OVERVIEW

AI Agent Governance
with InspectModular

Your AI agents are taking real actions — writing code, accessing databases, calling APIs, making decisions. InspectModular is the governance layer between your organization and every agent action.

Start free →EU AI Act guide →Set policy rules →

Three Pillars of AI Governance

🛡

Policy Enforcement

Define what AI agents are and are not allowed to do. Rules use pattern matching on any action — file operations, API calls, database writes, deployments.

  • Block rules — action is refused immediately
  • Warn rules — action is allowed but flagged
  • Pattern matching: "delete *", "access prod *", "deploy"
  • Rules apply to all agents in your workspace instantly
📋

Audit Trail

Every AI agent action is logged automatically — who did what, when, and with what outcome. No additional code required when connected via MCP.

  • Automatic — no code changes needed
  • Tamper-resistant — append-only storage
  • Full context: agent, action, target, outcome, timestamp
  • Searchable and exportable for regulatory evidence
🧠

Organizational Memory

Agents accumulate organizational knowledge across sessions. Context, decisions, and handoffs persist — so every agent starts informed, not cold.

  • Persistent facts across all agent sessions
  • Session handoffs — agents pick up where last one stopped
  • Searchable memory — agents find relevant context fast
  • Trust scoring — human facts weighted higher than agent facts

Regulatory Coverage

InspectModular's audit trail and policy enforcement directly address requirements from five major AI and data regulations.

EU AI Act
Article 12
Automatic logging of AI events throughout system lifetime
InspectModular audit trail logs every agent action automatically
SOC2 Type II
CC6.1, CC6.2
Logical access controls and audit logging for systems
Policy rules control access; audit trail proves it
HIPAA
Security Rule
Audit controls for systems accessing PHI
Every agent accessing patient data is logged and governed
GDPR
Article 22
Human oversight for automated decision-making
Policy enforcement + blocking ensures human oversight gates
ISO 42001
Clause 6-8
AI risk management and governance documentation
Policy rules and audit trail form the governance documentation
Deep dive: EU AI Act Article 12 compliance guide →

How It Works

🖥 For IDE agents
  1. Paste MCP JSON config into your IDE
  2. Agent calls primer_agent_brief at session start
  3. Agent calls primer_agent_check before risky actions
  4. Every call is logged to your audit trail
🐍 For Python/Node.js pipelines
  1. Install the InspectModular SDK or use REST API
  2. Call /api/v1/brief at pipeline start
  3. Call /api/v1/policy/check before sensitive operations
  4. All calls logged to the same unified audit trail

Frequently Asked Questions

What is AI agent governance?
AI agent governance is the practice of defining, enforcing, and auditing what AI agents are permitted to do in an organization. It involves: policy (rules defining allowed actions), enforcement (blocking violations), and audit trail (logging every action for compliance).
Why do companies need AI agent governance?
Three reasons: (1) Regulatory compliance — EU AI Act, SOC2, HIPAA, GDPR require audit trails and oversight for AI systems; (2) Risk management — AI agents can take destructive actions without guardrails; (3) Organizational trust — teams need to know what their AI agents are doing.
How does Primer policy enforcement work?
You define rules in the Policy page. Each rule has a pattern (e.g. "delete *", "access production"), an action (block or warn), and a severity. When an AI agent calls primer_agent_check before performing an action, Primer evaluates the action against all rules and returns allow, warn, or block.
Does governance work without IDEs — in Python pipelines?
Yes. Python and Node.js pipelines use the InspectModular REST API (/api/v1/policy/check, /api/v1/brief, etc.) or the InspectModular SDK. All calls log to the same unified audit trail as IDE agents. One dashboard governs everything.

Start governing your AI agents today

Free plan. No CLI. Connect via MCP in under 60 seconds. Policy rules and audit trail included.

Start free →EU AI Act guide →