Claude Managed Agents add an auto permission policy for agent and MCP tool calls, dated September 10, 2026

Claude's platform release notes, dated September 10, 2026, record a new permission policy for Claude Managed Agents: "Claude Managed Agents permission policies now include auto: the server evaluates each agent or MCP tool call and runs it, denies it, or pauses for your approval. agent.tool_use and agent.mcp_tool_use events report how each call was evaluated in an evaluation field alongside evaluated_permission."

Claude's permission-policies documentation places auto alongside the existing always_allow and always_ask policies, available to both the agent toolset and MCP toolsets, with no toolset using it by default. Under auto, a call runs "when the server determines that the call is safe," is denied when "the server evaluates the call as high-risk," or pauses for approval "when the server reaches no determination." Each agent.tool_use and agent.mcp_tool_use event now carries evaluated_permission ("allow", "ask", or "deny") and, in most cases, an evaluation object naming the policy that produced that outcome, with a reason_code attached when the outcome is ask or deny.