---
title: "Workday Agent Passport: Audit and Stop AI Agents — Kiweely"
description: "How Workday Agent Passport security stamps, Agent System Users (ASUs), audit trail reports, and Agent Management Hub deactivation govern AI agents."
url: https://kiweely.com/guides/workday-agent-passport-governance-explained
---

# What Is Workday Agent Passport and How Do Admins Audit or Stop an AI Agent?

Workday Agent Passport shows verified security stamps, assessors, and grant dates on each agent profile in Agent Management Hub. To audit or stop an AI agent, inspect View AI Agent User Audit Trail and click Deactivate Agent or disable its Agent System User OAuth 2.0 client.

-   **Checked** against Workday on November 13
-   Applies to **Workday AI**
-   6 min read
-   Published November 13
-   By Kiweely

**Workday Agent Passport** is the compliance and trust profile built into [Workday’s Agent System of Record (ASOR)](https://kiweely.com/guides/workday-agent-system-of-record-security) that displays verifiable **security stamps**, independent assessors, and grant dates for Workday-built, partner, and third-party AI agents. While Agent Passport proves how an agent was vetted before deployment, runtime auditing and emergency shutoff rely on **Agent System User (ASU)** identities, the **View AI Agent User Audit Trail** report, and **Deactivate Agent** (or OAuth 2.0 client disablement) inside **Agent Management Hub**.

Security and internal audit teams often conflate pre-deployment attestation with runtime enforcement. Understanding how Agent Passport pairs with ASU execution modes, **Agent Gateway**, and Workday’s compliance reports gives you a repeatable governance playbook before any agent touches HR or financial records.

## What Does Workday Agent Passport Actually Verify?

When an enterprise evaluates a Workday-built agent, a certified Partner agent, or an external third-party agent, vendor questionnaires rarely prove whether the agent was tested against prompt injection, data exfiltration, or model drift. **Agent Passport** surfaces structured security attestations directly inside your tenant so security administrators do not have to chase external PDFs.

To inspect an agent’s passport:

1.  Confirm **Agent System of Record** is set up in the tenant and your unconstrained admin group holds permissions on the *Agent Management Hub* domain.
2.  Open the **Agent Management Hub** report and select the **Agent Registry** tab.
3.  Click the registered agent’s name to open its profile, then select the **Passport** tab.
4.  Review the **Stamps** table, which lists each **Stamp Name**, **Description**, **Associated Assessors** (Workday alongside independent security partners such as Cisco AI Defense), and the **Date Granted** against industry AI security benchmarks (such as OWASP LLM Top 10, NIST AI RMF, and MITRE ATLAS). Click any stamp name to drill into its underlying compliance scope.

What Agent Passport does **not** do on its own is restrict which worker records or business processes the agent can touch inside your tenant. Runtime least-privilege is enforced by Workday’s **Agent Security** framework and **Agent System Users**.

## How Agent System Users (ASUs) Separate Delegate vs. Ambient Execution

Unlike traditional integrations that share broad Integration System Users (ISUs), Workday stores a dedicated **Agent System User (ASU)** identity in the Object Management System (OMS) for every configured agent. An agent can hold up to **two separate ASUs**, each with its own dedicated OAuth 2.0 client and key-pair pointer in ASOR, because interactive chat and background automation carry fundamentally different risk profiles:

| Governance Dimension | Delegate Mode | Ambient Mode |
| --- | --- | --- |
|  **Trigger** | Human worker prompts the agent interactively (for example, submitting PTO). | Event trigger or schedule fires in the background without human interaction. |
|  **Identity & ASU** | Dedicated **Delegated ASU** paired with the invoking human user’s session. | Dedicated **Ambient ASU** acting strictly as itself. |
|  **OAuth 2.0 Flow & Lifetime** | Authorization Code / On-Behalf-Of flow. Up to 60 minutes for Workday-built agents; 4 hours (24-hour refresh) for external agents (not yet customizable). | JWT Bearer assertion signed with the X.509 key pair registered during ambient skill setup. |
|  **Authorization Check** |  **Intersection** of the skill’s **Available To** interaction policy AND the human user’s domain, BP, row, and field permissions. | Evaluated solely against the auto-generated **Ambient Agent Security Group (ASG)** assigned to domain and BP policies. |
|  **Audit Trail Signature** | Logs the agent ASU as **By User** and the human requester as **On Behalf Of User**. | Logs only the ambient agent ASU as **By User**. |

Proxies and Delegations Are Blocked on ASUs

Workday blocks both **Start Proxy** and business-process delegation across agent interactions and ASUs. Auditors never have to untangle “proxy acting as manager invoking an agent on behalf of an employee” chains in production logs.

## Which Workday Reports Audit AI Agent Inputs, Outputs, and Accounts?

Granting an auditor *Agent Management Hub* alone is not enough to inspect forensic logs. Assign unconstrained audit security groups across four specific domains to expose the full compliance surface:

-   **View AI Agent User Audit Trail** (secured to *Reports: AI Agent Security* in the *System* functional area): Your primary forensic ledger showing every transaction executed by an ASU, cleanly separating autonomous ambient runs (*By User* only) from human-prompted delegate runs (*By User* + *On Behalf Of User*).
-   **All AI Agent Accounts** (secured to *Workday Accounts* in the *System* functional area; also visible in **All Workday Accounts**): Lists every generated ASU in the tenant so security admins can verify which ASUs are enabled vs. disabled.
-   **Input and Output Report For Registered Agents**, **Agent Input and Output Request Details**, and **Generate Agent Input and Output Report** (secured to *Agent Compliance* in *Agent System of Record*; available for Workday-built agents): Captures conversational input/output details over a rolling 30-day data source window (`Agent Input and Output Details`) and links directly to Workday Account related actions for audit trails.
-   **View Security for Agent Skill** (secured to *Manage: Agents* and *Security Administration*): Audits every API tool inside a skill, flags `Config Required` or `Review Required` mismatches, and provides **Test My Setup** to trace every security policy pathway granting a user access to that tool.

For [external and partner agents](https://kiweely.com/guides/workday-custom-agents-a2a-external-agents), Workday also requires all API calls (REST, SOAP, Graph, WQL, and RaaS) to route through regional **Workday Agent Gateway** endpoints (such as `https://us.agent.workday.com` or `https://eu.agent.workday.com`) with a unique `wd-agent-tenant-alias` HTTP header:

Fails

```
// Calling standard tenant web services with a shared integration account for an external AI agent:
GET https://<tenant-host>/ccx/service/<tenant>/Human_Resources/v43.0
// Bypasses Agent Gateway rate guardrails, ASOR session analytics, and ASU By-User / On-Behalf-Of separation.
```

Works

```
// Routing external agent traffic through the regional Workday Agent Gateway with ASU OAuth credentials:
GET https://us.agent.workday.com/api/common/v1/workers
wd-agent-tenant-alias: acme-prod
// Enforces ASU scopes, built-in rate limits, and centralized ASOR compliance logging.
```

## How Do Admins Throttle, Stop, or Permanently Kill a Misbehaving AI Agent?

Depending on whether an agent is hallucinating on one tool, burning unexpected [Flex Credits](https://kiweely.com/guides/workday-platform-consumption-console), or suffering a suspected credential compromise, Workday gives administrators four graduated kill switches:

1.  **Disable a single tool (blast radius: one API action):** For supported Workday-built agents like Self-Service Agent, open **Agent Registry > Skills > Configure (Tools)** and toggle off the offending tool while leaving the rest of the skill online. Takes up to 5 to 10 minutes to apply to new chat sessions.
2.  **Disable a skill or shrink `Available To` (blast radius: one capability or population):** Edit the agent in **Configure Agent**, toggle the skill’s **Status** off or remove security groups from **Available To** (for ambient skills, remove the **Ambient Agent Security Group** from the domain/BP policy and run **Activate Pending Security Policy Changes**).
3.  **Deactivate the agent (blast radius: entire agent in the tenant):** Open **Agent Management Hub > Agent Registry**, select the agent, and click **Deactivate Agent** (requires *Manage: Agents* and *Setup: Agents*). Workday hides the agent from users, stops Flex Credit burn in that tenant, and blocks **Agent Gateway** requests within **1 minute**, while preserving history so you can reactivate later or **Archive** it permanently.
4.  **Emergency OAuth 2.0 client kill switch (for suspected credential compromise):** Workday cannot detect stolen external OAuth credentials on your behalf, and external agent OAuth access grants remain valid for up to **4 hours** (`14,400` seconds). If you suspect a replay attack or leaked external client credential, do not stop at **Deactivate Agent**. Immediately disable the affected **OAuth 2.0 Client** in Workday security settings to sever replay attempts inside the remaining lifetime window.

Before activating any Workday-built or external agent, use [Kiweely Vine](https://kiweely.com/vine) to trace which existing domain and business process policies your target security groups already hold, surfacing hidden over-permissions across live tenant metadata before an ASU ever executes a write ([read our security architecture](https://kiweely.com/security)).

## Frequently asked questions

### Where do I find Workday Agent Passport in my tenant?

Ensure Agent System of Record is enabled and your security group has access to the Agent Management Hub domain. Open the Agent Management Hub report, select the Agent Registry tab, click an agent's name, and select the Passport tab to inspect its Stamps table.

### How does Workday distinguish autonomous agent actions from human-initiated actions in audit logs?

Each agent receives up to two Agent System User (ASU) accounts in the Workday Object Management System. In View AI Agent User Audit Trail, interactive delegate-mode actions record the agent ASU as By User and the human worker as On Behalf Of User, whereas scheduled ambient-mode actions record only the ambient ASU as By User.

### How fast does deactivating an AI agent cut off external API calls?

Clicking Deactivate Agent in Agent Management Hub hides the agent immediately in Workday and propagates to Workday Agent Gateway within up to 1 minute. If you suspect compromised OAuth credentials on an external agent whose 4-hour access grant is still live, immediately disable the affected OAuth 2.0 client as well.

### Does a Workday Sandbox refresh wipe out previous AI agent usage history?

No. Workday preserves previous weeks of agent usage activity across Sandbox refreshes, and changing data center regions is the primary action that resets analytics history in Agent Management Hub.

## Sources checked

-   [View Agent Passport Security for Agents](https://doc.workday.com/admin-guide/en-us/workday-ai/agent-security/view-agent-passport-security-for-agents.html)
-   [Agent Security Framework and Execution Modes](https://doc.workday.com/admin-guide/en-us/workday-ai/agent-security/concept--agent-security.html)
-   [About Workday Agents and Agent System of Record](https://doc.workday.com/admin-guide/en-us/workday-ai/agents/setup-considerations--agent-system-of-record.html)
-   [FAQ: Agent Security](https://doc.workday.com/admin-guide/en-us/workday-ai/agent-security/faq--agent-security.html)
-   [Concept: Workday Agent Gateway](https://doc.workday.com/admin-guide/en-us/workday-ai/agents/workday-agent-gateway-for-partner-agents.html)

## Related guides

[All guides](https://kiweely.com/guides)

-   [### What Workday Admins Need to Configure Before Turning On Sana Self-Service Agent Before activating Sana Self-Service Agent, run Set Up AI Admin Security Permissions, enable Agent System of Record, and add All Users to Graph API Descriptor. Keep the Guide skill enabled and align every skill’s Available To groups with underlying tool security policies.](https://kiweely.com/guides/workday-sana-self-service-agent-admin-setup)
-   [### How to Audit Over-Permissive Workday Security Groups Before Turning On AI Agents Delegate AI agents inherit the union of an invoking user’s security groups, while ambient agents act through dedicated Ambient Agent Security Groups. Before activating any agent, audit unconstrained User-Based groups, stale role assignments, and View Security for Agent Skill pathways.](https://kiweely.com/guides/workday-security-audit-before-enabling-ai-agents)
-   [### Connecting a Workday custom agent to an outside agent platform with A2A Over A2A, an outside platform calls your Workday custom agent, not the other way round. You register a third-party agent definition in ASOR that uses the custom agent as a tool, activate it, fetch its agent card and give the outside platform the OAuth details with PKCE.](https://kiweely.com/guides/workday-custom-agents-a2a-external-agents)
-   [### Workday Agent System of Record (ASOR): An Admin’s Security and Setup Checklist Workday’s Agent System of Record (ASOR) governs both Workday-built and custom agents through the Agent Management Hub. Here is how ASUs differ from ISUs, how delegate and ambient execution evaluate security, and the exact steps an administrator takes to register, secure, and audit an agent.](https://kiweely.com/guides/workday-agent-system-of-record-security)
-   [### Securing Workday Custom Agents: Permissions, Approvals and Audit A Workday custom agent acts for a person and is checked against both that person’s security and the agent’s own skills. How to check that boundary, add approvals and audit agent work.](https://kiweely.com/guides/securing-workday-custom-agents)

### More guides on Custom agents and Workday AI

[All 91 guides](https://kiweely.com/guides#custom-agents)

-   [What Workday’s Deployment Agent Can and Cannot Do in Your Tenant](https://kiweely.com/guides/workday-deployment-agent-tenant-inspection)
-   [Calling External Glean and Sana Agents From Workday Extend Without Hitting 408 Request Timeouts](https://kiweely.com/guides/workday-extend-external-agents-glean-sana-408-timeout)
-   [AI Gateway Document Intelligence (/ml/docint/v1/ids): Converting Workday Drive Base64 to Binary, Resume Layout Pitfalls, and Fixing Opt-In Errors](https://kiweely.com/guides/workday-document-intelligence-base64-binary-opt-in-errors)
-   [Why a Workday Custom Agent Has No Credentials Tab in Agent Management Hub, and How to Expose It Through A2A](https://kiweely.com/guides/workday-custom-agent-missing-credentials-tab-a2a)
-   [Passing Large CSV File Attachments From a Workday Custom Agent Flow to Orchestrate Without Hitting Context Limits](https://kiweely.com/guides/workday-custom-agent-csv-file-attachment-orchestrate)
-   [Preventing Duplicate Initiate\_Change\_Job Business Process Events When a Workday Custom Agent Calls a Write Tool](https://kiweely.com/guides/workday-custom-agent-duplicate-initiate-change-job)
-   [Why Custom Agent Actions Fail With S22 Permission Denied or Task Not Authorized in Agent Chat](https://kiweely.com/guides/workday-custom-agent-actions-permission-denied-tool-wids)
-   [Fixing invalid\_request OAuth Token Exchange Errors When Connecting Microsoft Copilot Studio to Workday A2A](https://kiweely.com/guides/workday-a2a-copilot-studio-oauth-invalid-request)

### From the blog on Custom agents and Workday AI

[All posts](https://kiweely.com/blog)

-   [Does Workday have an official MCP server, who can use it, and how do you connect Claude or ChatGPT?](https://kiweely.com/blog/workday-mcp-server-connect-external-agent)
-   [Adding an agent to a Workday Extend app you already have](https://kiweely.com/blog/agentify-existing-workday-extend-apps)
-   [How Independent Workday Consultants Deliver Like Full Implementation Teams Using AI Agents](https://kiweely.com/blog/how-independent-workday-consultants-compete-using-ai)

## Let Workday check the work.

Kiweely writes WQL and builds orchestrations in your own tenant, and Workday, not Kiweely, says whether they passed.

[Start for free](https://kiweely.com/download) [See how it works](https://kiweely.com/vine/how-it-works)

## Get the next guide in your inbox.

One email a month with new guides and one Extend tip. Nothing else.
