Anthropic adds budget controls, regional reasoning and adviser model features to Claude Managed Agents

Anthropic adds budget controls, regional reasoning and adviser model features to Claude Managed Agents

N
News Editor
2026-08-08 06:34:36
Anthropic updated Claude this week with four additions to Claude Managed Agents aimed at enterprise use. The changes cover session-level budget controls, regional reasoning controls, automatic loading of Skills modules from user repositories, and a new adviser model option. Users can now cap spending for an agent session, with the system set to trigger an event and pause execution once that limit is reached; work can resume after the budget is raised. Anthropic also introduced location controls for where Claude Managed Agents run. Jobs can be scheduled across globally available resources at standard pricing, while restricting execution to the United States costs 1.1x the standard rate to meet regional deployment requirements. The update also lets agents automatically load Skills modules stored in user repositories. In addition, users can configure a stronger model as an adviser, allowing the active agent to call on it during a session for a second-opinion analysis intended to improve performance on more complex tasks.

Anthropic updated Claude this week, rolling out four new features for Claude Managed Agents focused on enterprise control, deployment flexibility and task execution.

The first addition is session budget control. Users can now set a spending cap for an agent session to manage costs more precisely. Once a session reaches that limit, the system triggers an event and pauses execution. The task can resume after the user raises the budget.

The second change is regional reasoning control. Users can choose where Claude Managed Agents run, scheduling workloads across globally available resources at standard pricing. If execution is restricted to the United States to meet regional deployment requirements, the cost is 1.1x the standard price.

The third update adds support for automatically loading Skills modules from a user repository.

The fourth is an adviser model feature. Users can configure a more powerful model as an adviser, allowing the task-running agent to call that model during a session for a second-opinion analysis, with the stated goal of improving the quality of handling more complex tasks.

This article was originally published by Bit.Fan. For more cryptocurrency news and market insights, visit www.bit.fan.
760

Disclaimer:

The market information, project data, and third-party content displayed on this platform are for industry information sharing only and do not constitute any form of investment advice or return commitment.

Cryptocurrency trading carries high risks. Users should fully assess their risk tolerance and make independent decisions. All profits, losses, and legal responsibilities are borne by the users themselves.