ServiceNow announced on October 15, 2026 that every AI agent gets a cryptographically signed credential called an AI Passport, and that employees can now build their own agents inside EmployeeWorks under admin-defined limits.
The AI Passport is an authority layer that gives companies runtime control over what AI agents and Model Context Protocol servers are allowed to do at every point in a multi-agent workflow. Each agent, whether it is ServiceNow’s own or a third-party agent, carries a credential that states its identity, permitted actions, accessible systems, and accountable owner. Agents check that credential as the work runs. NVIDIA OpenShell is named as a design and launch partner. ServiceNow built the system on open specifications so other governance products, identity providers, or enforcement points can issue and verify the credentials.
When an agent stays inside its allowed scope, it proceeds. When it drifts, the Super Intelligence Control Tower can redirect it, restrict it, reconnect it with the proper context, or trigger a kill switch. Every action is audited. The credentials are issued by the Control Tower and grounded in the company’s Configuration Management Database, governance, risk, and compliance policies, and workflow approvals. ServiceNow expects AI Passport, shadow AI detection, and AI Cost Management to be available in the fourth quarter of 2026.
The same announcement expands the Super Intelligence Control Tower with shadow AI detection that finds unapproved agents on the network, more than 400 connectors for discovery across platforms, access mapping tied to actual entitlements, continuous observability of agent traces, cost governance that tracks spending by user, department, and runtime, and value intelligence that connects consumption to cost, quality, and realized user value. AI Cost Management inside IT Asset Management brings frontier model, embedded software-as-a-service AI, hyperscaler, and data-center GPU costs into one view.
On the employee side, ServiceNow updated EmployeeWorks so people can build the agents their jobs demand. These tools run on a third-generation employee-centric AI harness that combines reasoning, memory, and tool interoperability, and they are in controlled availability today. Productivity Assistants are specialized assistants for contextual or persona-based work. Non-coding employees can build or tailor them under admin guardrails, and pre-built ones start with HR Service Delivery so managers get help on talent, growth, and hiring. Productivity Routines let employees create agents for predictable, repeatable work that runs in the background. ToolSmith is an agentic developer environment where developers describe the automation they want in natural language and receive an enterprise-grade tool in minutes. The Conversations API surfaces the assistants, routines, and ToolSmith capabilities inside the other applications employees already use.
For companies, the practical change is the ability to move agents from pilots into production with limits that are checked as the work happens rather than only reviewed afterward. A company can set a customer-service agent to process refunds under a fixed amount and block anything higher for human review. Shadow detection and the cost tools give one place to see what is running and what it is spending. Employees can create useful assistants and routines without every request landing on IT, while the same control plane limits what those agents can do. Companies already running ServiceNow can ground the credentials in the configuration and policy data they already maintain.
Several other platforms offer pieces of agent governance or employee agent-building. Microsoft has Agent 365, Entra Agent ID, and Copilot Studio. Amazon has Bedrock AgentCore with its identity, gateway, and policy features. Salesforce has Agentforce with its trust layer and fabric. IBM has the agentic control plane inside watsonx Orchestrate. Google has agent identity and gateway tools. Security vendors such as Palo Alto, Cisco, Arthur AI, and others focus on discovery, policy enforcement, or kill switches. Specialized identity tools include Keycard and various agent-passport offerings. ServiceNow’s approach ties the credential and the runtime checks to its configuration management database and workflow system, uses open specifications, and pairs the control with employee self-service building under the same harness.
I covered the earlier expansion of the AI Control Tower at Knowledge 2026. This announcement adds the verifiable credential that travels with the agent and the employee building tools that sit under the same rules.
Ask your ServiceNow team when AI Passport will be available for the agents you already run, and whether the Control Tower data that issues the credential is current. Confirm which third-party agents and MCP servers can receive and check the passport. Test the EmployeeWorks building tools on one routine your team repeats before controlled availability widens.
Official announcement: ServiceNow advances its operational harness for enterprise AI with AI Passport, Business Wire, October 15, 2026.
Related: ServiceNow Knowledge 2026: From Workflows to an Autonomous Workforce.
