GitHub improved Copilot CLI's delegation logic to avoid unnecessary handoffs between agents, cutting tool failures by 23% and wait times by 3–5% in production testing.

What happened

GitHub released an update to Copilot CLI (version 1.0.42 or later) that makes the main agent more selective about when to delegate tasks to helper agents. The change rolled out to 100% of production traffic. In A/B testing, it reduced tool failures per session by 23%, search tool failures by 27%, and edit tool failures by 18%, while lowering total user wait time by 5% at P95 and 3% at P75.

Why it matters

Delegating work to subagents adds coordination overhead and wait time. The update teaches Copilot to handle focused, narrow tasks directly and reserve delegation for work that truly benefits from independent exploration or parallel execution. This reduces friction in the agent loop without requiring developers to manage new settings—the improvement happens behind the scenes.

What to watch

The update is available now by running the /update command in your terminal to version 1.0.42 or later. The gains came primarily from avoiding unnecessary subagent paths and reducing subagent workload per user, with no quality regression reported.