Codenotary AgentMon 3: An adaptive runtime security tool with integrated organisational history
Codenotary has this month detailed its AgentMon 3 offering.
This is an enterprise AI security offering with adaptive runtime security policies.
What makes a runtime security policy adaptive?
It’s one that is capable of continuously evolving as AI agents operate across an organisation by learning from customer-specific workflows, observed behavioural patterns and new, changing and/or emerging threats.
AI agents evolve constantly through new prompts, model upgrades, tool integrations, memory expansion and workflow changes.
AgentMon currently observes and analyses (before then securing) in excess of five million AI agent interactions every day.
From the channel end of the spectrum, AgentMon is now available through AWS Marketplace.
Production agent peculiarities
Codenotary claims its operational scale has provided it with real-world insight into how autonomous AI agents behave in production, enabling the company to build adaptive security capabilities based on actual enterprise deployments rather than laboratory simulations.
“As organisations 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, CEO and co-founder, Codenotary.
AgentMon 3’s adaptive behavioural model is built for autonomous AI agents. Every action an AI agent performs continuously shapes a live behavioural baseline, allowing organisations to distinguish normal operations from risky or anomalous behaviour in real time.
“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 organisations to secure AI at enterprise scale without creating an unsustainable operational burden,” added CEO Bar.
AgentMon replaces manual policy writing with dynamically generated, self-refining security policies based on how each organisation actually uses AI. It creates what the company calls “living policies” that adapt automatically across teams, roles, agents and workflows, creating precise security baselines unique to each environment.
Integrated organisational history
Unlike signature-based security products, AgentMon continuously integrates organisational history, evolving agent capabilities, changing workflows and emerging AI threat intelligence into its policy engine. As enterprise AI deployments mature, AgentMon’s adaptive security policies mature with them.
AgentMon also correlates intent with impact. Detection is based on observed file access, network activity, credential use, process execution and system connections – not on the agent’s own self-report.
This makes AgentMon resilient to prompt obfuscation, multilingual attacks and evasion techniques that bypass text-only security filters.
