New · The guide “Mastering Claude at work”, free. Download it →

Last reviewed:

Running Claude Code autonomously: delegate the loop, keep control

In autonomous mode, you no longer drive every step of Claude Code: you set a stop condition once, and an evaluator checks it after each turn until the work is genuinely finished. Four levers make this loop automatic: /goal, /loop, /schedule and Stop hooks.

Claude Code can work two ways: as an assistant you watch at every step, or as an autonomous agent you delegate a « work, check, repeat » loop to until a stop condition is met. This dossier details the four mechanisms that make that autonomy operational, and where to keep control.

4levers turn Claude Code from an assistant you babysit into an agent that finishes the job.

WYP.agency — practical guide, verified edition · July 2026

01From assistant to autonomous agent

Most developers use Claude Code as an assistant they must watch constantly. There is another way to work: you set a condition once, and Claude runs on its own until it is met.

In the usual flow, you are the loop: you write a prompt, Claude replies, you review, you re-run, twenty times. In the autonomous flow, you set a stop condition, an evaluator checks it after each turn, and if it is not done, Claude goes again. You define the goal once.

WYP take

This is not a convenience setting. It is a different relationship to the tool: instead of driving every step, you delegate the loop « work, check, repeat ». The question is no longer « how do I do this », but « what am I willing to delegate ».

02The four levers

Four mechanisms make the loop automatic. Each is defined by its stop condition — what decides the work is done.

1

/goal — runs until a condition is true

You describe the target; Claude chains turns until an evaluator confirms it is reached. Best when « done » is objectively checkable: the tests pass, the page loads.

2

/loop — re-runs on a regular interval

Claude is re-run in a loop until you stop it or the work is done. Useful to watch a changing state: a deployment, a queue.

3

/schedule — at fixed times, in the background

The task fires on a schedule, with no open session. For the recurring: a report every morning, a watch every night.

4

Stop hooks — scriptable control of a turn's end

A script decides whether Claude may stop or must continue. The stop condition becomes programmable — the door to genuinely bespoke autonomy.

03Choosing the right lever

/goala checkable condition, you want the resultsession required
/loopa changing state, you monitorsession required
/schedulerecurring, in the backgroundno session
Stop hooksa bespoke stop rulescope-dependent

04Keeping control

Delegating the loop does not mean letting go. Useful autonomy is bounded: a clear stop condition, a limited scope, and a point where a human signs off on what commits.

WYP take

The right question is not « does it work? » but « which decision am I willing to delegate, and which requires my signature? » Autonomy is framed before it is launched.

Get the guide as a PDF

The four levers to run Claude Code autonomously — /goal, /loop, /schedule, Stop hooks — each command cross-checked against the official documentation.

PDF · 12 pages · FR / EN / ES · delivered in under a minute

See also

Further reading

Documentation Claude Code — Anthropic (documentation officielle)

Sources

  1. WYP.agency — Guide « Faire tourner Claude Code en autonomie » (édition vérifiée, juillet 2026). https://wyp.agency/glossaire/claude-code-autonomie/ (accessed 2026-07-09)
  2. Anthropic — Documentation Claude Code. https://docs.claude.com/en/docs/claude-code (accessed 2026-07-09)

← Back to glossary

Address copied