Shaping what Azure SRE Agent does: Tool Permissions and Hooks (opens in new tab)

When an AI agent runs against production, the first question every security team asks is "What can it do, who decided it could, and what stops it from doing something it should not." Azure SRE Agent reached general availability in March. Since then, teams inside Microsoft and customers running it against real production workloads have asked for the same thing: finer-grained controls over what the agent can do on its own and a clear answer to who governs each call that reaches a tool. Today at...