Codenotary AgentMon 3

AgentMon 3 is Codenotary's runtime security and monitoring platform for AI agents. It automatically discovers coding assistants such as Cursor, Claude Code and Codex, as well as custom workers. It then learns a behavioural baseline for each agent and blocks risky actions. Security, compliance and AI-operations teams get one place to see agent activity, cost, and policy violations across the enterprise.

Codenotary AgentMon is a runtime oversight platform for AI agents. It launched on 31 March 2026 as enterprise monitoring for agentic networks, and AgentMon 3 followed on 8 July 2026 together with an AWS Marketplace listing. The product automatically discovers the agents running across an organisation, including Cursor, Claude Code, Codex, OpenCode, Amp and custom LangChain-style workers. It tracks each agent's health, communication paths, model choice, token consumption, inference latency, file access, secrets handling and data-access patterns. According to the AWS Marketplace listing, telemetry is collected through eBPF and OpenTelemetry, and the vendor stresses that no proprietary SDK is required. Policies are written in Rego and evaluated at the agent's pre-tool-use hook. A simulation mode shows how often a new rule would have fired against the last 30 days of traces before anyone switches it to enforcement. The headline change in version 3 is adaptive runtime policy: every action updates a live behavioural baseline, and each decision is weighed against agent identity, permissions, history, data sensitivity, prior human approvals and threat intelligence. Codenotary claims this cuts policy maintenance by up to 80%. Every runtime decision is written to Codenotary's immutable ledger, which gives security teams a verifiable audit trail for compliance and forensics. Because AgentMon watches actual behaviour, it keeps detecting risky actions even when an agent's own permission prompts are bypassed or disabled. Codenotary reported more than 5 million monitored interactions per day in July 2026, and its product page cited more than 7 million in September 2026. AgentMon competes in the young AI-agent security and observability category, which is growing quickly.

Category

Governance & Security

Pricing

Contact for pricing

Target Market

CISOs, CIOs, Security Teams, AI Operations Teams, Compliance Leaders

Deployment

Self-hosted

Headquarters

Houston, USA

Customers

Codenotary reports hundreds of customers worldwide, including banks, governments and defence organisations (company-wide figure, July 2026)

• ✓
  Automatic agent discovery

  Finds every AI agent running in the company, including Cursor, Claude Code, Codex, OpenCode, Amp and custom workers, which exposes shadow AI usage.

• ✓
  Adaptive runtime policies (v3)

  Builds a live behavioural baseline from each agent's actions and evolves the policies automatically, so teams maintain far fewer hand-written allow-lists.

• ✓
  Rego policy engine with simulation mode

  Evaluates Rego rules at the agent's pre-tool-use hook and shows how often a rule would have fired over 30 days of traces before it is enforced.

• ✓
  Immutable audit ledger

  Records every runtime decision cryptographically in Codenotary's tamper-proof ledger, which gives auditors and investigators a verifiable history of agent behaviour.

• ✓
  Cost and runaway-agent controls

  Tracks spend in real time by agent, model and team, sets budgets and limits, and alerts on looping or overspending agents before costs escalate.

• ✓
  Secret redaction and risky-action detection

  Scrubs leaked secrets from telemetry before storage and flags dangerous shell commands, sensitive file writes, exposed API keys and hostile inputs.

• ✓
  OpenTelemetry-native telemetry with SIEM export

  Collects data via OpenTelemetry and eBPF without proprietary SDKs, and exports audit trails to SIEM tools and alerts to Slack or PagerDuty.

• •
  Governing AI coding assistants

  A security team inventories every Claude Code, Cursor and Codex instance on developer machines and blocks dangerous shell commands or writes to sensitive files at runtime.

• •
  Compliance evidence for regulated industries

  Banks, government and defence organisations keep a cryptographically verifiable ledger of each agent decision, which they can hand to auditors or use in incident forensics.

• •
  Controlling agent spend

  A platform team attributes token spend to each agent, model and team, sets budgets, and gets alerts when an agent loops or overspends, which enables chargebacks.

• •
  Safe rollout of new agent policies

  Security engineers simulate a new Rego rule against thirty days of recorded traces to see exactly how often it would fire before switching it to enforcement.

• •
  Securing multi-agent production systems

  Operations teams watch communication paths, latency and data-access patterns across orchestration frameworks and cloud services to catch anomalous behaviour or possible data leakage early.

Codenotary publishes no list price for AgentMon. Its public pricing page covers other Codenotary products, not AgentMon. The AWS Marketplace listing is bring-your-own-licence, so buyers pay AWS only for infrastructure and license the software directly from Codenotary. The product page offers a free trial button, and the AWS listing says evaluation licences are available from sales on request.

This page was written from 10 sources, 7 on domains other than codenotary.com.

1. 1.codenotary.com — agentmonvendor
2. 2.codenotary.com — ai agent risk monitoringvendor
3. 3.codenotary.com — pricingvendor
4. 4.aws.amazon.com — prodview bnwla36m4kpoo
5. 5.helpnetsecurity.com — codenotary agentmon agentic ai
6. 6.helpnetsecurity.com — codenotary launches ai security platform that learns from ai
7. 7.securitybrief.news — codenotary launches agentmon 3 for ai agent security
8. 8.efficientlyconnected.com — codenotary agentmon 3 ai agent runtime security
9. 9.tfir.io — codenotary agentmon ai agent monitoring
10. 10.enterprisemanagement.com — codenotary agentmon 3 demonstrates why ai runtime governance