Skip to content

Anthropic

Managed Agents get session budgets and an advisor model

Written by an AI model from the linked sources and checked against them by a second model. How we work

Video: x.com

ClaudeDevs posted on Twitter about four Claude Managed Agents updates shipped over the past week. Three of them:

• Sessions can now have a budget. Hit the limit and the session pauses with a budget_reached event; raise the budget and it picks up where it left off • Skills come from connected repos. If they’re already sitting in .claude/skills/ for Claude Code, the session loads them at startup automatically • Any agent can be given an advisor, a stronger model for a second opinion mid-session. It takes one line in the agent list

The previous batch, on July 22, brought per-agent effort levels, webhooks, and the ability to attach up to 50 events when creating a session. A session can carry up to 500 skills.

Anthropic

Claude Code turns auto mode on by default August 14

Starting August 14, Claude Code sessions on Pro, Max, and Team start in auto mode, where a classifier screens every tool call: it blocked 89% of dangerous commands in Anthropic’s tests, versus 13.6% caught by manual approval. Testers approve 97% of permission prompts.

No comments yet. Be the first.

Join the conversation

Sign in to leave a comment. It appears after moderation.