What is AI agent governance?

AI agent governance is the set of controls that decide which business systems AI agents can reach, what they can do there and what gets recorded. In StackOne that means connector and account access, member-based Permission Policies, optional prompt injection scanning, audit logs and your choice of deployment.

How does StackOne govern what an AI agent can do?

StackOne governs agents in layers. A connector profile sets the available actions and who can link accounts. Account permissions decide who can use each connected account. Permission Policies restrict tools, protect fields from changes or mask returned data on member-based calls. Prompt Injection Guard scans tool responses when enabled, and logs record each decision.

Does StackOne detect shadow AI?

No. StackOne governs the calls agents make through StackOne. It doesn't scan devices for unmanaged AI tools or inspect agents that connect directly to a business system. Routing agents through StackOne is what brings them under these access controls, policies and logs.

Which calls do StackOne's governance controls apply to?

StackOne's governance controls apply to calls made through StackOne. Action scoping applies to every call through a connector profile. Account permissions and Permission Policies apply when the call carries a member identity. API-key calls skip those member checks, so limit their actions through the connector profile and protect the keys.

How is StackOne Governance different from Permission Policies?

Permission Policies are one part of StackOne governance. They control which tools, inputs and returned data are allowed on member-based requests. Governance also covers connector and account access, identity and directory sync, prompt injection scanning, audit logs and deployment.

Does the AI model ever see a business-system API key in StackOne?

Not through StackOne. StackOne supplies the business-system credentials through managed authentication when it executes an action. Your agent doesn't need those credentials in its prompt or tool arguments, so the model never has to handle them.

AI agent governance. Govern what your teams' AI agents can do.

Control how your teams use AI agents across business systems, with security policies and audit logs in one place.

Start Free Book Demo

Same request, three people

Get Vendor Bill INV-2041NetSuite

Lena OrtizFinance
ChatGPT

Allowed Returns the vendor bill.

Decided byConnector accessFinance shares this NetSuite account
Priya ShahProcurement
Claude

Masked Returns the bill with bank details hidden.

Decided byPermission PoliciesProtect bank details · Enforce
Sam RiveraOperations
Claude Code

Refused Sam left yesterday. The call is refused.

Decided bySSO and SCIMDeactivated in Okta

AI agent governance case study. Nine business systems connected in under two weeks.

A US medical research organisation rolled out agents across IT and finance, with access set per connector and per account.

Medical research laboratory — **IT leader**US medical research organisation

Signing dayFirst live agent query

Within two weeksNine business systems connected

Configure access. Set access for teams and individual users.

Choose who can connect each system, who can use shared connections, and what their agents can read or change.

What you shareWho receives accessWhat that access permits

Connector profile

Connection setup and available actions

Users or groups

Can connect an account

The new connection uses this profile's action selection.

Connected account

An existing connection to a business system

Users or groups

Can use that connection

Account Members can run its available actions.

Policies apply when an agent acts for a member

Assign further restrictions to users or groups with account access.

Explore Permission Policies

Restrict actionsProtect writesMask returned data

Sync members and groups from your directory. Manage their access grants in StackOne.

Explore SSO and SCIM

Choose accounts and available actions

Define actions through connector profiles. Share profiles for new connections, or grant access to accounts already connected.

Explore Connector Access

Set rules for member-based calls

Use Permission Policies to restrict tools, prevent changes to protected fields or mask returned data. API-key calls use separate access arrangements.

Explore Permission Policies

Authentication and response protection. Protect credentials and check tool responses.

StackOne authenticates calls to business systems. Enable Prompt Injection Guard to inspect returned content for prompt injection.

Authenticating an action

AI agent

StackOne cloud

Uses stored credentials

NetSuite

Get Vendor BillAuthenticated

Keep business-system credentials out of agent prompts.

Explore Agent Authentication

Inspecting a tool response

Vendor details

CompanyAster Industrial

Account statusActive

Prompt Injection GuardNo prompt injection detected

Clean

Prompt Injection Guard95.9%

Meta Prompt Guard v271.9%

Balanced accuracy on the Jayavibhav benchmark (65,415 samples). Meta Prompt Guard v2 is the next best model measured.

The leading prompt injection model checks responses before your agent reads them.

Explore Prompt Injection Guard

Observability and audit. See what agents did and who changed the setup.

Find action outcomes, changes to StackOne and security activity in the relevant logs.

Action logs

21 Sep 2026UTC

Recent actions Policies Defender

User / Agent	System / Action	Outcome / Checks
Taylor Brooksvia Copilot	SAP S/4HANA 12:41:33	200 2 Clean
Noah Patelvia Claude Code	Jamf Pro 12:42:10	200 2 Clean
Alex Morganvia Claude	Workday 12:43:51	200 3 Clean
Lena Ortizvia Copilot	Salesforce 12:44:06	200 2 Clean
Priya Shahvia Claude Code	Xero 12:45:19	403 3 Not run
Sam Riveravia Claude Code	ServiceNow 12:46:32	201 2 Clean
Alex Morganvia Claude Code	Workday 12:47:58	200 3 Clean

Workday

Get Worker

21 Sep · 12:47:58 UTC

200 OK Clean

User / Agent: Alex Morgan via Claude Code
Account: People Operations

Policies evaluated3

Protect personal dataMasked

EnforceMatchedResponse filtering

Field groupPersonal datapii

Personal data masked before reaching the agent.

Keep salaries unchangedNo match

Review worker lookupsMonitor onlyWould deny

2 other policiesNo match · Would deny

Worker lookup completed

Use logs in your own reporting and investigations

Retrieve action and platform records through the Logs API and forward them through your pipeline.

Explore Observability & Audit

Deployment and compliance. Choose the deployment that fits your requirements.

Compare managed cloud with regional projects, a dedicated account and self-hosting.

Shared cloud

Encryption keyOurs or yours

StackOne hosts and operates the service.

Choose a region for each project

Dedicated account

Encryption keyOurs or yours

An environment dedicated to your organization.

Agree region and operating arrangements

Self-hosted

Encryption keyYours

Your team operates StackOne in your environment.

Separate licence and support agreement

Plan residency and operations

Discuss where StackOne runs, who operates it and how encryption keys are managed.

Explore Deployment

SOC 2 Type II security review. Evidence for your security review.

"As a company rooted in security assurance, we needed a partner sharing our priorities. StackOne's approach gave us confidence to scale without compromise."

Faraz YaghoutiSenior Director of Product, Drata

Read the Drata story

SOC 2 Type II

Request the report, and review subprocessors and security information, in our trust centre.

Visit the trust centre

FAQ. Questions about AI agent governance

See how StackOne fits your AI rollout.

Talk through the systems your teams use and the controls your rollout needs.

Book a demo Start free