← Full glossary
Agent inventory
gobernanzaagent registry
An agent inventory is the up-to-date register of the AI agents operating in an organization: what they are, who is accountable for them, which tools and data they touch and with which credentials.
Implementa's formulation. It is the starting point of any agent governance: what is not inventoried cannot be assigned an owner, bounded or audited. It also includes unofficial ('shadow') agents found through sensors, telemetry or review of tools and MCP servers. Identity-governance vendors offer it as a discovery feature; the inventory remains an organizational practice, not just a product function.
How it differs from
- Model registry
- A model registry lists AI models; an agent inventory lists agents with their permissions and owners.
- CMDB
- A CMDB inventories IT assets; an agent inventory adds autonomy, tools and a business owner.
Examples
- A living sheet listing each agent, its owner, tools, credentials and last review date.
- A discovery scan that finds an agent built by a marketing team with CRM access unknown to IT.
FAQ
- Where to start?
- With agents that touch sensitive data or take actions with financial effect, and by assigning an owner to each.
- Does regulation require it?
- No text calls it that in general; the EU AI Act requires risk management and logging for high-risk systems, and an inventory is a practical way to meet that.
Related terms
Sources & further reading
- Help Net Security, SailPoint Agentic Fabric · 2026-10-06
- KuppingerCole, Omada gains runtime agent governance · 2026-09-25