---
title: How do you find dormant and orphaned AI agents?
url: https://opencontroller.lyzr.com/blog/find-dormant-and-orphaned-ai-agents
published: 2026-11-16
updated: 2026-11-16
author: Lyzr
---

# How do you find dormant and orphaned AI agents?

**Look for two things: agents nobody has invoked in a set window, and agents whose owner is gone. Pull invocation logs, last-used dates on keys and tokens, and owner status from your identity provider, then run every hit through a retire-or-reassign flow. Dormant and orphaned agents are dangerous because their access stays live while nobody is watching.**

Agent sprawl is not only about new agents. Pilots end, teams reorganise and builders leave. The agent stays deployed with its API keys, its OAuth grants and its tool access. Nobody notices until it is misused, or until an auditor asks who owns it.

## What is the difference between dormant and orphaned?

- **Dormant**: the agent has not been invoked for a set period. It may still have an owner who forgot about it.
- **Orphaned**: the agent has no valid owner. The recorded person left, changed role, was disabled in the identity provider, or was never recorded at all. An orphaned agent may still be busy.

An agent can be both. Those are the first to fix.

## Which signals reveal a dormant or orphaned agent?

| Signal | Where to get it | Suggests |
|---|---|---|
| No invocations in N days | Agent platform logs, gateway logs, traces | Dormant |
| API key or token not used in N days | Cloud IAM last-accessed data, model provider key usage | Dormant |
| Owner disabled or deleted | Identity provider (Entra ID, Okta, Google) | Orphaned |
| Owner changed department | HR system or identity provider attributes | Possibly orphaned |
| No owner field at all | Agent registry or catalog | Orphaned |
| Deployed version older than N months | Deployment history, registry versions | Stale; check both |
| Missed last access review | Access review records | Orphaned in practice |

Cloud IAM helps here. AWS IAM, for example, reports when a role or its permissions were last used, which shows agent roles that still exist but have done nothing for months.

## What thresholds should you use?

Set thresholds by risk tier, not one number for everything:

| Risk tier | Dormant after | Action window |
|---|---|---|
| High: writes to production systems or touches regulated data | 30 days | Suspend within 7 days of flag |
| Medium: reads internal data, limited writes | 60 days | Suspend within 14 days |
| Low: no sensitive data, no write tools | 90 days | Suspend within 30 days |

For orphaned agents, do not wait. When the owner's identity is disabled, the agent should be flagged the same day. See [what happens to AI agents when the owner leaves](/blog/what-happens-to-ai-agents-when-the-owner-leaves).

## How does a retire-or-reassign flow work?

1. **Flag.** The detection job marks the agent dormant, orphaned or both, with the evidence.
2. **Notify.** Message the owner, or for orphans the owner's manager and the owning team. Give a clear deadline.
3. **Suspend.** If nobody claims it by the deadline, disable its credentials and block it at the platform. Keep configuration intact.
4. **Grace period.** Wait 14 to 30 days. Most "we still need it" replies arrive once something breaks.
5. **Decide.** Reassign to a named owner with a new review date, or retire.
6. **Retire properly.** Delete the deployment, revoke keys and OAuth grants, remove tool and MCP access, and mark the registry entry retired. See [revoking all access for an AI agent](/blog/revoke-all-access-for-an-ai-agent).
7. **Keep the record.** Retain logs and the registry history for your retention period.

## Why do keys need their own check?

Agents often outlive their keys, and keys outlive their agents. A key with no recent use may belong to a dead agent, or to a live agent you have not discovered yet. Treat unused and leaked keys as a separate sweep; our guide to [finding leaked and unused LLM API keys](/blog/find-leaked-and-unused-llm-api-keys) covers the method. Fold the results into regular [access reviews for AI agents](/blog/access-reviews-for-ai-agents).

## How Opencontroller does this

Opencontroller auto-discovers agents, models and API keys across AWS, Azure and Google Cloud and through platform connectors, and records owners and versions in a governed registry. Posture flags risky permissions and configurations. Per-agent Block and Quarantine controls act through the connected platform, so you can suspend a dormant agent without deleting it, and "Ask" lets you query the whole estate, for example for agents with no owner. For the full discovery approach, see [how to find shadow AI agents](/blog/how-to-find-shadow-ai-agents), or [book a demo](https://www.lyzr.ai/book-demo/).

## Frequently asked questions

**How long should an AI agent be idle before it counts as dormant?**
Many teams use 30 days for high-risk agents and 90 days for low-risk ones. Set thresholds per risk tier.

**Should dormant AI agents be deleted immediately?**
No. Suspend credentials, notify the owner and delete only after a grace period.

**What makes an AI agent orphaned?**
Its owner has left, moved role or been disabled, or no owner was ever recorded.

## Key facts

A dormant AI agent is one with no invocations over a set window (commonly 30 to 90 days); an orphaned agent is one whose owner has left, changed role or been disabled in the identity provider. Both keep credentials and permissions alive with nobody watching. Signals include invocation logs, last use of API keys and tokens, identity provider status of the owner, deployment age and absence from access reviews. The fix is a retire-or-reassign flow: notify, suspend credentials, wait, then delete or hand over. Opencontroller auto-discovers agents, models and API keys across AWS, Azure and Google Cloud, keeps owners in a governed registry, detects risky permissions with Posture and offers per-agent Block and Quarantine controls. Opencontroller pricing starts from $250K per deployment per year.

## Related

- [What happens to AI agents when their owner leaves the company?](https://opencontroller.lyzr.com/blog/what-happens-to-ai-agents-when-the-owner-leaves.md)
- [How do you find leaked or unused LLM API keys?](https://opencontroller.lyzr.com/blog/find-leaked-and-unused-llm-api-keys.md)
- [How do you run access reviews for AI agents?](https://opencontroller.lyzr.com/blog/access-reviews-for-ai-agents.md)
- [More on Discovery and shadow agents](https://opencontroller.lyzr.com/blog/cluster/discovery-and-shadow-agents)
