Ping extends privileged access controls to Claude AI agents

Ping Identity is applying privileged access controls to actions that Claude and other locally operated AI agents perform on employees’ behalf.

The company introduces Enterprise Personal Agent Access through PingOne Privilege, with Claude Desktop and Claude Code identified as early use cases.

The capability places access controls between an employee’s agent and the enterprise systems it attempts to use.

For instance, when an agent requests access to a managed resource, Ping associates the session with the employee and device that originated it.

The enterprise can then evaluate the request against its access policies, limit the available permissions, require approval, or reject the action.

The approach allows an agent to receive temporary access to an approved tool or service without storing a reusable password, API key, or other long-lived credential inside the application.

The technical documentation identifies two integration paths for personal agents: an inline gateway for Model Context Protocol (MCP) connections and controls for activity involving Git repositories.

An MCP request routed through the Ping gateway can be evaluated, restricted, and recorded before it reaches the protected service. A Git operation using the configured privilege workflow can similarly be linked to the employee and subjected to policy.

This provides an access control layer around selected agent actions, but it does not give Ping control over agents’ reasoning. The protection depends partly on whether the organization can keep sensitive resources within the controlled route.

Last year, Ping was developing Identity for AI to discover agents, create identities for them, and govern their access. By March 2026, Ping had made elements of the portfolio generally available.

The company is also positioning its control plane to discover agents, grant temporary privileges, and link AI-assisted software commits to the employee who made them.

Ping also cites its participation in Anthropic’s Project Glasswing. Anthropic describes Glasswing as a defensive security research initiative that involves AI-assisted vulnerability research and software protection.

Ping says that Enterprise Personal Agent Access is available and has entered pilot programs with several large enterprises.

Article Topics

AI agents  |  Claude  |  digital identity  |  identity access management (IAM)  |  Ping Identity