AnthropicapiAuto Mode and Session Viewer for Claude Managed Agents
Two developer updates to Claude Managed Agents: Claude can now judge each tool call itself, and you can watch a running agent from your terminal or browser.
Details
- Setting the permission policy to auto has Claude review each tool call and MCP tool call against the intent in your messages, then run it, deny it, or pause to ask you
- Each tool-use event now carries an evaluation field with the decision and a short reason, such as flagging a potentially destructive operation
- The new ant beta:sessions connect command attaches your terminal to a running session so you can follow it live, send messages, and approve or reject pending tool calls
- Adding --web opens the same session in a browser view served from your own machine
- The session viewer is in beta in the ant CLI. Both updates require API access to Claude Managed Agents
Best for
Letting a hosted agent run without approving every tool call by hand, while still stopping risky actions
When to use it
Your fixed allow and deny lists either block useful work or let too much through.
Who benefits
Developers and technical teams running agents on the Claude Platform
Join the newsletter to get ahead with AI.
I share strategies, new modules, free guides, and tools to get you better at using AI in your work.
Leading a team? See team programs →