Claude Code update lets users set the effort level of each sub-agent

Anthropic released a Claude Code update on the 6th of October 2026, according to the Claude Code changelog on GitHub. Claude can now run each sub-agent at the effort level the user asks for.

Claude Code is Anthropic's AI assistant for writing software.

Why it matters

A sub-agent is a helper that Claude starts for one side task, such as searching many files, Anthropic's documentation says. Each helper uses the same usage limits as the main chat.

Low effort is "faster and cheaper for straightforward tasks", Anthropic's guide says. High effort "provides deeper reasoning for complex problems".

Users can now give small side tasks low effort and hard ones high effort.

Key facts

The guide lists five levels: low, medium, high, xhigh and max. Claude starts at medium unless the user picks another, it says.

At higher levels, Claude tested more edge cases and checked more of its own work, the guide says. At lower levels it answered sooner.

Before this update, the effort level could only be set in the file that defines a helper, the documentation says.

The same update adds one-step plugin installs, the changelog says. It lists 92 changes, 64 of them fixes.

What's next

Anthropic released two more updates, on the 7th and the 8th of October, according to GitHub. On the 8th of October, the sub-agent documentation still described only the older setting.