
What happens to access control when your users are AI agents moving faster than any human, coordinating with each other, and… not always telling the truth about what they're doing? James Governor sits down with Bobby DeSimone, founder & CEO of Pomerium to discuss the new frontiers of access control. Pomerium is an open source, self-hostable, access control platform named for the ancient boundary around Rome. Bobby and James dig into why authorizing every action matters more than identity alone, and why static OAuth scopes are a poor fit for MCP. Bobby also makes a contrarian case against some of the AI knower herd: MCP will beat CLIs, he argues, because MCP makes it much easier to keep credentials out of the agent sandbox. He walks through Pomerium's architecture, a context-aware reverse proxy built on Envoy. He then argues the industry needs new primitives, including "temporal access control" (James thinks this is super important!). This control will need to be able to evaluate an agent's recent history and behaviour.
Show notes: https://redmonk.com/videos/bobby-desimone/
Chapters:
00:00 Introduction to AI Gateways and Pomerium
02:21 Understanding Pomerium: The Founding Story
03:58 The Importance of Context in Access Control
05:49 Balancing Action and Identity in Security
09:46 Technical Overview of Pomerium
12:37 Policy Setting and Enforcement in Pomerium
13:50 The State of Standards in Agent Security
15:07 CLI vs. MCP: Security Considerations
17:50 The Evolution of Security Models
20:21 Bobby's Journey into Security
23:01 The Journey to Pomerium
27:36 Understanding Pomerium's Architecture
29:51 User Experience and Security
32:24 Transitioning from Human to Agent Security
34:43 The Future of Security in Agent Collaboration
More episodes from "The MonkCast"



Don't miss an episode of “The MonkCast” and subscribe to it in the GetPodcast app.








