HYPR announced the launch of HYPR AgentPass, an identity security platform designed to address the risks associated with autonomous AI agents. The solution links each AI agent to a designated human supervisor and evaluates requests against identity and security policies before actions are executed. High-risk activities are routed to supervisors for passkey-signed approval, while authorized users can stop an agent from executing its next request through a centralized console or mobile device. The platform helps enterprises strengthen AI agent accountability, governance, and human oversight.
As autonomous agents take on daily tasks across corporate networks, most run on borrowed credentials such as shared service accounts or personal developer API keys. Because nothing links an agent’s actions to the person it is acting for, security teams cannot verify which agent initiated an action or on whose behalf, enforce permissions, or stop an unauthorized request.
With Gartner predicting that through 2028 over 50% of enterprise AI initiatives will halt due to unresolved agentic identity challenges, organizations must close this visibility gap before scaling deployment.
Existing approaches address parts of the problem. AI gateways, while an important component, route agent traffic, and prompt-inspection tools look for dangerous inputs. AgentPass answers a different question before every model or tool call: should this agent, acting for this person, be allowed to do this? It also covers agents a network gateway never sees, including coding agents and their sub-agents running on developer workstations.
“Due to very real security concerns, every enterprise is lacking the confidence to enable agents to act autonomously, which is the only way significant ROI can be achieved,” said Bojan Simic, CEO and co-founder of HYPR. “Security teams cannot protect what they cannot govern. AgentPass brings the same verifiable identity assurance we provide for human workforces directly to agents, ensuring that every agent action is authorized, bound by policy, and tied to an accountable individual. Our mission is to provide the necessary agentic controls that gives businesses the confidence to execute entire workflows with agents, securely”
Core Capabilities of HYPR AgentPass
AgentPass delivers a unified control plane spanning developer endpoints, a central agent gateway, and real-time administrator controls:
-
Agentic Accountability on Every Action: AgentPass binds each agent to a named human supervisor who signs in through the organization’s identity provider. All agent connections run through a single governed gateway, where agents start with zero access, see only the tools permitted by their supervisor’s role, and never hold the credentials to those tools; those credentials stay in the gateway and are issued as short-lived, single-tool session tokens. Because access follows the supervisor’s role, a role change updates the agent’s access, and offboarding cuts it off on the agent’s next request.
-
Endpoint and Mobile Supervision of Agents: Discovers AI tools and coding agents, including Claude Code, Codex, Cursor, and others, on enrolled macOS and Windows workstations, attributed to a user and device. AgentPass then routes the supported tools an administrator chooses through its gateway and blocks the bypass, keeping local agents on a governed path. This extends governance to coding agents and the sub-agents they launch. A mobile app allows humans remote supervision and full visibility into what their agents are up to.
-
Real-Time Guardrails: Checks every agent request against policy at the gateway, outside the model, before it runs, so a prompt injection cannot argue its way past a rule. Administrators can write policies in plain English, test them on live traffic in “Observe” mode, then enforce them to allow, block, or hold actions for human approval, such as letting an agent draft an email but requiring sign-off before it sends, or limiting which recipients and domains it can reach.
-
Passkey Approvals & Instant Kill Switch: Routes high-risk actions and new agent connections to a supervisor’s mobile device, a separate device from the one the agent runs on, for FIDO2 passkey approval. This ensures the human signs the exact action requested on an independent device. Authorized users can disable a rogue agent from the console or the HYPR mobile app, and its next request is refused everywhere, including in sessions already running.
-
Continuous Supervision & Pantheon of AI Watchers: Relying on static rules and deterministic policy engines is insufficient when securing agents that act probabilistically. AgentPass provides out-of-the-box evaluator agents that continuously monitor and supervise all agents running across the enterprise. Whether enforcing a secure SDLC for developers agents or ensuring cyber red team agents remain within their sandbox, AgentPass provides a library of watchers to secure every agentic action with the appropriate context.
HYPR AgentPass operates seamlessly across existing enterprise identity infrastructure. Organizations can deploy it at their own pace, starting with visibility into agent activity and baseline governance, then adding and enforcing policies as they are ready. AgentPass connects to the identity and security products customers already run and also integrates directly with the HYPR Identity Assurance Platform.
For more details, visit https://www.hypr.com/
Related News:
HYPR Research Finds 98% of HR Executives Encounter Candidate Fraud
TELUS Digital Poll Finds 46% of AI Users Bend the Truth to Reach Human Customer Service Agents