Identity for AI Agents

Build secure identity for agents that accept requests, call services, act on behalf of users, and delegate work to other agents. Every agent action should answer: who is acting, what can it access, and on whose authority?

What You Can Build

Protected Agents

Require callers to present valid tokens before invoking agent capabilities.

Agent Credentials

Delegated Access

Let agents act on behalf of users with consent, scope intersection, and actor context.

Multi-Agent Workflows

Downscope tokens across agent chains while preserving traceability from the original user or caller.

When to Use This Pattern

Use this pattern when an agent calls protected services, exposes capabilities to clients, acts on behalf of users, stores credentials, delegates to other agents, or must produce an audit trail.

How Agent Identity Works

In this model:

Each agent is a first-class identity with its own credentials and assigned permissions.
Callers authenticate before invoking protected agent capabilities.
ThunderID issues tokens that represent the agent, the user, or both depending on the flow.
Services validate tokens, check scopes, and enforce audience restrictions.
Delegated and multi-agent flows preserve actor context so downstream services can audit who initiated each action.

Building Blocks of the AI Agent Identity Journey

Use these building blocks to design agent identity across inbound requests, outbound service calls, and multi-agent delegation. Select a block to see what it solves, where it appears in the journey, which capabilities it uses, and which guide to start with.

Solution Patterns

Choose the token and delegation pattern before implementation. Agent identity design depends on whether the agent acts as itself, acts for a user, receives background approval, or delegates work to other agents.

Client Credentials

Use this when an agent acts autonomously as its own identity to call protected services.

Authorization Code with OBO

Use this when an agent receives user context and needs delegated tokens for downstream services.

Backchannel Authorization

Use this when a background agent needs out-of-band user approval before acting.

Token Exchange

Use this when one agent delegates to another agent or needs a narrower token for a specific service.

Review the full pattern guide: Solution Patterns.

Implementation Paths

Protect Your Agent

Require callers to present valid tokens before invoking agent capabilities.

Start with: Try It Out - Protect Your Agent

Connect to Services

Let agents call APIs, MCP servers, tools, and internal services with the right token flow.

Start with: Try It Out - Connect to Services

Secure Multi-Agent Workflows

Downscope tokens as one agent delegates work to another and preserve the original actor context.

Start with: Token Exchange

Apply Scope Intersection

Keep delegated tokens within the overlap of user authority and agent permissions.

Start with: Scope Granularity

Manage Credential Lifecycle

Issue, rotate, expire, revoke, and audit credentials for each agent identity.

Start with: Credential Lifecycle

Lock Tokens to Audiences

Bind tokens to the service they are meant to call so they cannot be replayed elsewhere.

Start with: Audience Locking

Audit Delegation Chains

Trace every agent action and delegation hop back to the original caller or user.

Start with: Delegation Chain Preservation

Recommended Starting Point

Start by choosing the token pattern, then register the agent and define the resource scopes it needs. After that, protect the agent's inbound API and configure the outbound service flow.

B2C Overview

Use this when public-facing users sign themselves up and manage their own accounts.

B2B SaaS Identity

Use this when organizations, tenants, delegated admins, and enterprise customers shape the identity model.

What You Can Build​