Skip to content
AI Framework

Govern AI agents on CrewAI

Add governance to CrewAI multi-agent workflows. The CrewAI pack enforces policy on task delegation, requires approvals for inter-agent handoffs, and provides audit trails across your entire agent crew.

What this pack does

  • Policy enforcement on crew task delegation
  • Approval gates for inter-agent handoffs
  • Output safety on crew results
  • Cross-agent audit trail

Use cases

Require approval when agents delegate sensitive tasks

Enforce budget limits on crew execution

Audit the full delegation chain across crew agents

Quick setup

  1. 1Install the CrewAI pack: cordumctl pack install crewai
  2. 2Integrate the Cordum governance hook into your crew
  3. 3Define delegation and task-level policies
  4. 4Enable the pack and run your crew with governance

Frequently asked questions

How does Cordum govern a multi-agent crew?

You integrate the Cordum governance hook into your crew so that task delegation and inter-agent handoffs pass through policy. The Safety Kernel evaluates each step, returning Allow, Deny, or Require Approval before one agent hands work to another.

Can I require approval when agents delegate sensitive tasks?

Yes. Delegation-level policy lets you gate inter-agent handoffs, so a sensitive task returns a Require Approval decision before it is passed on. That prevents a crew from silently routing high-risk work between agents without oversight.

Can I cap how much a crew run is allowed to do?

Yes. You can enforce budget and task-level limits on crew execution, so a run that would exceed your constraints is denied or held for approval. This keeps an autonomous crew from expanding its own scope beyond what you intended.

How do we trace decisions across the whole crew?

Cordum maintains a cross-agent audit trail that captures the full delegation chain and the governance decision at each handoff. You can reconstruct exactly which agent delegated what to whom and on whose approval.

Ready to govern CrewAI?

Other integrations