AI & MLVideo
ABAC and end-to-end auditing: how to secure MCP-connected AI agents?
In the final part of our series, Tomasz Janaszka, Solutions Architect at CodiLime, moves beyond simple role-based permissions to show how Attribute-Based Access Control (ABAC) and unified auditing can prevent unauthorized network access and provide total visibility.
This video demonstrates how to implement dynamic security rules that adapt to context, such as location, tenant, or maintenance windows.
We cover the advanced layers of the security stack, specifically:
- Why "Read/Write" roles aren't enough. Tomasz explains how to use Open Policy Agent (OPA) to evaluate JWT tokens against real-time device inventory. Learn to restrict access based on attributes like allowed_sites, allowed_envs, and allowed_tenants.
- A live demo of a dynamic policy that can instantly lock out non-admin users during critical maintenance, without restarting servers or reconfiguring agents.
- How to overcome the challenge of 8+ independent components logging separately. Tomasz introduces Correlation IDs that propagate through the entire chain – from the user’s natural language query to the final CLI execution.
- A look at the logging architecture using Syslog-ng, Promtail, Loki, and Grafana to create a searchable, end-to-end audit trail.








