How Many Credentials Should Your AI Agent Have? Zero. — Jim Clark, Docker

Why watch

Jim Clark, principal software engineer at Docker, explains a security-first approach to AI agent architecture using MCP gateways and sandbox isolation. The talk demonstrates through a newsroom workflow example (researcher, fact-checker, publisher sandboxes) and a coding agent example how to limit what context and tools agents can access, ultimately achieving zero credentials inside sandboxes by using Okta's Cross App Access for identity and authorization. This is essential viewing for AI engineers building secure, scalable agent systems.