Claude Code adds dynamic loops: AI automatically determines monitoring cadence, evolving from polling to event-driven

Anthropic engineer Noah Zweben announced that Claude Code has added dynamic looping functionality on April 10. When users run the

/loop

command and do not specify an interval time, Claude automatically determines when to check next based on the nature of the task—possibly even using the Monitor tool directly to bypass polling, fundamentally changing how AI agents perform monitoring.

From fixed polling to intelligent scheduling

In the past, the

/loop

feature required users to manually set the check interval (for example, running once every 5 minutes). The issue is that a CI test that may finish in 2 minutes and a build that takes 45 minutes are forced to use the same polling frequency—the former wastes waiting time, while the latter wastes meaningless check次数.

Dynamic looping lets Claude decide the pacing itself. For example, when running

/loop check CI on my PR

, Claude observes the status of the CI. If it detects that the build has just started, it may set a longer waiting time; if it detects that the process is nearing completion, it shortens the check interval.

Monitor tool: from polling to event-driven

The more critical change is the integration of the Monitor tool. Community members have pointed out that this is essentially an architectural shift from cron (scheduled polling) to webhook (event-driven) systems. Claude can directly monitor a target (such as the output of a development server) and respond immediately when events occur, rather than repeatedly checking whether it is “done.”

This is in line with the Monitor Tool that was introduced earlier by Anthropic product lead trq212. trq212 previously suggested that developers use “start my dev server and use the MonitorTool to observe for errors,” so Claude keeps watching instead of performing intermittent queries.

Practical application scenarios

Dynamic looping is suitable for any task that needs to wait for uncertain durations:

Monitoring CI/CD pipeline completion status

Waiting until a deployment is finished to verify whether the service is operating normally

Observing error output from a development server

Tracking long-running data processing tasks

Currently,

/loop

runs locally, so the Claude Code session must keep running. There are already calls in the community to raise the three limits of /schedule (cloud scheduling) to support more automation scenarios.

Claude Code continues evolving into a developer operating system

Dynamic looping is part of Claude Code’s recent flurry of updates. In the past week alone, Anthropic also rolled out the /ultraplan web planning feature, the Advisor strategy, and Claude for Word—showing that Anthropic is turning Claude Code from a “code generation tool” into a complete developer agent operating system.

This article, “Claude Code Adds Dynamic Looping: AI Automatically Determines Monitoring Cadence, Evolving from Polling to Event-Driven,” first appeared on 链新闻 ABMedia.