Agent Identity

On-Behalf-Of Access

On-behalf-of access describes when an AI agent invokes a tool or system using access granted in the context of a human user's task, where the agent is acting as a delegate of that user.

Why this matters for AI agents

On-behalf-of patterns are common in agentic systems—users ask agents to perform tasks that require access to their data. Without enforcement, agents can silently expand beyond the user's intended scope.

How AgntID relates

AgntID enforces on-behalf-of access by binding the agent's execution context to the initiating user's task scope—agents cannot exceed the access appropriate for the original user intent.

The static IAM gap

OAuth provides on-behalf-of tokens for API access but does not evaluate whether the agent using that token is invoking only the tools appropriate for the task that generated the token.

Related phrases

on behalf of accessdelegated agent accessuser-delegated agentimpersonated agent access

Related terms

Frequently asked questions

What is on-behalf-of access for AI agents?

On-behalf-of access describes when an AI agent invokes a tool or system using access granted in the context of a human user's task, where the agent is acting as a delegate of that user.

What is the risk of on-behalf-of access without enforcement?

Without enforcement, agents can silently expand beyond the user's intended scope—accessing data or taking actions the original user never authorized.

How does AgntID constrain on-behalf-of access?

AgntID enforces on-behalf-of access by binding the agent's execution context to the initiating user's task scope, ensuring agents cannot exceed the access appropriate for the original user intent.

Runtime Enforcement

Secure every agent tool call at execution time.

AgntID gives infrastructure teams scoped, ephemeral access control for AI agents without replacing IAM, MCP servers, tools, or agent frameworks.