Codenotary launches AgentMon 3 for AI agent security

Codenotary has launched AgentMon 3 for AI agent security, and the product is now available through AWS Marketplace.

The release adds adaptive runtime security policies designed to change as AI agents operate in customer environments. AgentMon now secures more than 5 million AI agent interactions each day across enterprise deployments, according to Codenotary.

The Houston-based company is positioning the latest version as a response to the growing use of autonomous software tools in coding, business automation and customer support. It argues that fixed security controls are becoming harder to maintain as AI agents evolve through new prompts, model updates, tool integrations and workflow changes.

AgentMon 3 builds behavioural baselines from observed activity and uses them to flag actions that appear risky or unusual. Codenotary says the system generates and refines policies based on how each organisation uses AI, rather than relying on manually written rules.

The approach can reduce policy maintenance work by up to 80%, the company said. The platform also monitors runtime behaviour independently of built-in permissions and allow-lists in AI tools, allowing it to detect high-risk actions even when those native controls are bypassed, misconfigured or disabled.

Runtime focus

The platform evaluates security decisions using contextual signals including agent identity, permissions, prior behaviour patterns, data sensitivity, requested resources, earlier human approvals and current threat intelligence. Codenotary says this is intended to improve detection while reducing false alarms.

Detection is based on observed activity such as file access, network traffic, credential use, process execution and system connections. In other words, it examines what an agent does at runtime rather than depending on the agent's own description of its intent.

All runtime decisions are recorded in what Codenotary describes as an immutable ledger, creating an audit trail for compliance reviews, investigations and forensic work. The product is aimed at organisations running coding assistants, autonomous software engineering agents, internal AI systems, orchestration frameworks, cloud services and multi-agent environments.

The AWS Marketplace listing expands the company's reach to customers that already buy and manage software through Amazon Web Services. Marketplace availability has become a common route for software vendors seeking to shorten procurement cycles in large organisations.

Codenotary says the scale of its installed base shaped the latest release. It describes AgentMon as one of the larger AI runtime security platforms in production and says real-world data from enterprise environments informed the adaptive policy model.

Security around AI agents has become a more prominent issue as companies give software systems broader permissions to access code repositories, internal tools, data stores and business processes. Those connections can make AI agents useful, but they also expand the range of actions that need oversight.

Traditional cybersecurity tools often depend on static signatures, pre-defined rules or text-based filters. Codenotary says those methods can struggle when autonomous systems frequently alter their behaviour or when attacks use obfuscated prompts, multiple languages or techniques designed to evade simple content checks.

Moshe Bar, Chief Executive Officer and Co-Founder of Codenotary, said existing approaches are under strain as AI use spreads across large organisations.

"As organizations rapidly deploy coding assistants, autonomous software engineering agents, business automation platforms, AI-powered customer support systems, and custom orchestration frameworks, traditional security models that are based on static allow-lists and manually maintained policies are proving increasingly inadequate," said Moshe Bar, Chief Executive Officer and Co-Founder of Codenotary.

He said the product is intended to adapt to customer-specific usage patterns while incorporating information about new threats.

"AgentMon 3 continuously learns from millions of real-world agent interactions while adapting to each customer's environment and incorporating intelligence from newly emerging threats. That enables organizations to secure AI at enterprise scale without creating an unsustainable operational burden," Bar said.

Codenotary says its customer base includes banks, governments and defence organisations, and that the platform is used by hundreds of customers worldwide. AgentMon 3 is available immediately worldwide and through AWS Marketplace.