Answer: Uber said it built an internal agent platform in early 2025 to let teams compose, deploy, and operate production-grade agents at scale, and enabled MCP (Model Context Protocol) support over existing service APIs. Uber also described updates to its identity and access stack to establish verifiable cryptographic identity for agents.
Uber said a key problem in agentic workflows is that execution context and attribution can be dropped across “agent hops,” making it hard to answer “who did what, when and why” for auditing, compliance, and trust. It outlined components including an Agent Registry as a source of truth, a Security Token Service that issues short-lived scoped JWT tokens per hop, and an “AI Agent Mesh” described as the communications data plane among agents and tools. The post frames this work as extending Uber’s Zero Trust Architecture to AI agents deployed as workloads, with token-based authentication for outbound calls such as to MCP tools.
Source: Uber