// AI Platform Drift

Claude Managed Agents adds auto permission mode and live terminal attachment

Claude Platform release notes · published 2026-09-11 03:26 UTC
WORTH KNOWING VERIFY_ONLY APIMODEL
WHAT CHANGEDClaude Managed Agents now supports an 'auto' permission policy mode where the server evaluates each tool call and either runs it, denies it, or pauses for approval. The agent.tool_use and agent.mcp_tool_use events now include an 'evaluation' field reporting how each call was evaluated. The ant CLI adds 'ant beta:sessions connect' command to attach a terminal to live agent sessions, with an optional --web flag to serve the session viewer locally.
WHY IT MATTERSIf you are running Claude Managed Agents with custom permission policies, the new auto mode changes how tool call authorization works—the server now makes evaluation decisions instead of requiring you to implement all logic client-side. The new evaluation field in events means your logging or monitoring code may need updates to capture this data. The CLI command provides a new debugging path that bypasses your application's session handling.
WHAT TO DOReview your agent permission policy configurations and decide whether to adopt auto mode or continue with existing policies. Update any event handlers that parse agent.tool_use or agent.mcp_tool_use events to handle the new evaluation field. Test the ant beta:sessions connect command in development to understand how it interacts with your session management.
Model-reported confidence 75% (not measured accuracy) Primary source ↗ written by anthropic/claude-sonnet-4.5
Show the reasoning — raw diff, triage verdict, confidence
RAW DIFF
+## Claude Platform release notes — September 10, 2026
+https://platform.claude.com/docs/en/release-notes/overview#september-10-2026
+Thu, 10 Sep 2026 00:00:00 GMT
+
+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. See Let the server evaluate each call with auto.
+The ant CLI adds ant beta:sessions connect, which attaches your terminal to a Claude Managed Agents session. You can follow the session live, send messages, and allow or deny tool calls that are waiting for approval. Pass --web to serve the Claude Console's session viewer locally and open the session there instead. See Connect to a Managed Agents session from your terminal.
HOW IT WAS JUDGED
Triage signals below are model-reported assertions, preserved verbatim; they are not independent arithmetic verification.
triage verdictMATERIAL at 0.92, needed 0.45
triage reasoningNew agent permission modes, API fields, and CLI commands affect how teams integrate and control Claude Managed Agents.
signalsnew feature: auto permission policy mode for Claude Managed Agents · new API field: evaluation field in agent.tool_use and agent.mcp_tool_use events · new CLI command: ant beta:sessions connect · new capability: terminal attachment to live agent sessions · new parameter: --web flag for local console serving
suppression—
triage modelanthropic/claude-haiku-4.5 ($0.00000)
writer modelanthropic/claude-sonnet-4.5 ($0.00000)