AI Agents

Tool calling is easy. Trust isn’t. Every new wave of infrastructure brings its own version of the “simpler protocol.” With AI agents, that moment arrived fast — the Universal Tool Calling Protocol (UTCP) has started making noise as the next big thing. Its promise sounds familiar: no wrapper servers, no middleware, no proxy hops. Agents can “just call” APIs, CLIs, or services directly using a JSON manual. Elegant. Minimal. Free of the heavy machinery that came with the Model Context Protocol (MCP). ...

If you missed my last post on what makes something an identity, start there — it sets the groundwork. This piece goes deeper: how to architect identity as the control plane for enterprises running cloud workloads and autonomous agents. This isn’t theory. It’s about production identity architectures that handle millions of authentications, thousands of microservices, and the new security challenges of AI agents. Whether you’re securing traditional enterprise apps, cloud-native services, or agent-driven workflows, the patterns here offer a roadmap — from where most organizations are today to where identity is heading. ...