Agent Review
The Agent Review is an enterprise governance system that ensures every AI agent is reviewed and governed by humans before it can operate in your organization. Think of it as a governance board for AI deployments.
What is the Agent Review?
The Agent Review is a human governance board that gates agent deployment. When enabled on an organization, agents cannot transition from draft to active status without passing a structured security review and receiving enough governance votes from reviewers.
Why does it exist?
AI agents can access sensitive data, call external APIs, browse the web, and execute workflows autonomously. The Agent Review ensures every agent has been evaluated for data safety, tool scope, autonomy limits, network access, compliance requirements, and prompt injection risks before it goes live.
How it works
- Admin enables Agent Review on the organization and adds members.
- Agent owner submits their agent with a 43-question security questionnaire covering 8 domains.
- Reviewers review the submission and vote to approve or reject.
- On governance, a guardrail profile is compiled from the answers and attached to the agent.
- At runtime, CleverGuard enforces the guardrail profile on every action the agent takes.
- If the agent changes after governance, drift detection blocks deployment until re-review.
Roles
Enables agent review, sets governance thresholds, adds/removes members.
Reviews submissions, votes to approve or reject, provides rationale.
Submits agents for review, responds to rejections, resubmits when needed.