Concepts
Understanding the following concepts helps you configure and manage agent identities in IBM Verify
AI agent
An AI agent is an autonomous software program that perceives its environment, makes decisions, and takes actions to achieve a specific goal. Agents can operate independently or on behalf of a user.
Agent identity
An agent identity is a purpose-built, non-human identity construct assigned to an AI agent. It is distinct from user accounts and application identities. Agent identities enable IBM Verify SaaS to authenticate agents, enforce access policies, and maintain a governance record of agent activity.
Onboarding an agent
Onboarding is the process of registering an AI agent with IBM Verify. An agent must be onboarded before it can authenticate or access resources. The onboarding process can be completed either through the IBM Verify SaaS UI or SCIM API and assigns the agent an identity and the credentials it uses to obtain tokens.
Agent lifecycle
An agent identity has the following states:
| State | Description |
|---|---|
| Default | It indicates that the activationDate and
expirationDate must be used to determine if the agent is
active |
| Active | The agent is registered and can authenticate and access resources |
| Suspended | The agent is suspended from IBM Verify SaaS |
| Under review | Indicates that the agent is not yet ready for use and is under review. The service provider determines how this status should be treated. For example, the service provider can treat this the same as an inactive agent, or use it to indicate that the agent is still active but under investigation, which can have a bearing on access decisions |
Token-based authentication
Agents do not use passwords. They authenticate by using tokens issued through standard OAuth 2.0 grant flows:
- JWT Bearer grant flow — The agent presents a signed JWT to obtain an identity token
- Client credentials grant flow — The agent authenticates by using a client ID and secret to obtain an access token
On-behalf-of (OBO) token
An OBO token allows an agent to act on behalf of a specific user. The token carries the user's identity context, ensuring the agent's actions are scoped to what that user is authorised to do. This prevents agents from having unrestricted access to all user data.
Privileged token
Not all agent actions are straightforward user-delegated operations. When an action requires elevated access, strict policy enforcement, and a clear audit trail, a privileged token provides a narrow, short-lived, policy-controlled grant. This ensures IBM Verify can authorise the specific operation without giving the agent broad or persistent access.
Agent tool integrations
IBM Verify SaaS supports agents that interact with resources through the following integration patterns:
- Direct tools — The agent connects directly to tools or services, such as a GitHub repository
- MCP tools — The agent connects to an MCP server that manages tool access by using OAuth
- IBM Secrets Vault — The agent retrieves secrets or credentials stored in IBM Secrets Vault
- Agent-to-Agent (A2A) — The agent communicates with another agent to delegate tasks, share context, or compose multi-agent workflows. Each participating agent authenticates independently and the requesting agent must be authorised to invoke the target agent