Akeyless has announced the general availability of Akeyless Agentic Runtime Authority, the real-time identity control layer for AI agent actions. It works on top of Akeyless SecretlessAI, a credential protection layer that keeps credentials out of AI agents and brokers access to enterprise systems. Runtime Authority adds the next layer of control, enforcing intent based access control restricting what agents actually do once they have access.
AI agents today do more than answer questions. They access databases and applications, execute workflows, and make decisions inside live production environments. Traditional access controls can determine whether an agent is allowed into a system, but not whether every action it takes once inside is legitimate.
For autonomous agents that can act unexpectedly or be manipulated, the impact can escalate at machine speed, as the July 2026 Hugging Face incident demonstrated. Runtime Authority closes that gap by evaluating agent actions in real time and blocking those that violate policy before they execute.
With general availability, Akeyless is expanding its AI agent ecosystem with new integrations for Claude Enterprise, OpenAI Codex, and Amazon Bedrock AgentCore, alongside new runtime control and investigation capabilities for production deployments. Runtime Authority runs on the same Akeyless Identity Security Platform that already secures over 220 billion machine identity interactions for Fortune 500 organizations, now applied to a new identity class: the AI agent.
“Our customers are the ones defining what security for agentic workflows should actually look like, and we build alongside them as trusted partners in that work,” said Oded Hareven, CEO, Akeyless.
“Take a coding agent pulling credentials to reach a cloud resource, or a sales executive trusting that an AI assistant will honor a ‘read-only’ permission on a sensitive SaaS application. Both come down to the same question: is there one place issuing, governing, and revoking that authority? Humans and machines will continue interacting with critical systems, so having a central point of access control and attribution only matters more over time.”
Runtime Authority: Enforce dynamic policy at the moment of action
Runtime Authority controls how legitimate access is used once an agent is connected, layering intent enforcement on top of the more traditional role- and attribute-based controls. It evaluates an agent’s objectives and actions in real time and blocks activity that violates policy or exceeds its assigned task before it executes. This is a powerful “kill switch” that revokes the authority of the agent to act in real-time when things go wrong. For example, an agent instructed to summarize sales data would be prevented from deleting a database or exfiltrating large quantities of data, even if its underlying permissions would otherwise allow it.
Securing AI agents from credential to action
SecretlessAI addresses the credential problem by keeping credentials out of AI agents entirely. Instead of giving an agent a password, token, or key that it can store, expose, or reuse, Akeyless brokers access through the Akeyless Gateway, which provisions a short-lived identity directly on the target system, even legacy and on-prem systems that use traditional identity methods. A compromised or prompt-injected agent has nothing to steal and nothing to leak.
Runtime Authority uses that same brokered access path to add control over what the agent can do once connected and prevents risky actions before they cause damage. Together, SecretlessAI and Runtime Authority protect the full path from credential to action.
Governance & visibility to agent action
Across both the credential and action layers, Akeyless gives security teams the visibility and control needed to monitor, investigate, and respond to agent activity. Every action is logged and traced back to the human, application, or agent that triggered it.
General availability adds new operational controls for production environments. The Agentic Access Dashboard provides a live view of active agent sessions, including the ability to terminate a session immediately, while new investigation capabilities show why a session was blocked. Enriched security events can also be forwarded to Splunk, Datadog, and Microsoft Sentinel for centralized monitoring and investigation.
Expanded AI agent ecosystem
The new integrations with Claude Enterprise, OpenAI Codex, and Amazon Bedrock AgentCore join existing support for a broad range of AI models, agent frameworks, and developer environments, including Google Gemini, xAI Grok, Claude Desktop, Cursor, GitHub Copilot, JetBrains IDEs, and n8n. This gives organizations a consistent way to apply SecretlessAI credential protection and Runtime Authority controls across the mix of AI platforms and tools used across the enterprise.
Agentic Runtime Authority works within the Akeyless Identity Security Platform that organizations already use for secrets management, certificate lifecycle management, privileged account security, and password management.