Practical overview
When it fits
Use Looper before running an agent task when the goal, definition of done, review criteria or stop conditions need to be made explicit and inspectable.
Who it is for
- Claude Code users designing repeatable agent workflows
- Developers who want explicit verification and review gates before an agent task runs
- Users preparing portable loop specifications for later execution
Documented uses
- Turn an automation idea into a structured goal, plan, verification and stopping workflow.
- Inspect a proposed loop in a terminal-friendly flow preview before finalizing it.
- Create portable YAML and compiled JSON specifications plus human-readable instructions and a Python runner script.
- Use a named loop pattern, such as code review, security scan, bug hunt or documentation sync, as a starting point.
Requirements & limits
The project is a public GitHub repository with an MIT license. The README documents installation for Windows PowerShell and macOS/Linux.
- Looper is a design layer, not itself the scheduled or persistent execution mechanism; its output can be handed to an execution route.
- The workflow can include programmatic checks, judge rubrics and human signoff, as well as revision caps and stop guards.
- The README documents both installer scripts and manual installation, and the helper compiler uses PyYAML in a private virtual environment.
- The repository identifies Python as its primary language and Claude Code as the target environment.
Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.
Install Looper and open its design workflow
Install the Claude Code skill and invoke its loop-design interview.
Before you start
- Claude Code
- Windows PowerShell, or macOS/Linux shell
- Internet access to retrieve the documented installer or GitHub repository
Steps
On Windows PowerShell, run the documented installer command: irm https://raw.githubusercontent.com/ksimback/looper/main/install.ps1 | iex. On macOS/Linux, run: curl -fsSL https://raw.githubusercontent.com/ksimback/looper/main/install.sh | bash.
Alternatively, follow the README's platform-specific manual installation commands to clone the repository into the Claude skills directory and copy the slash-command file into the commands directory.
In Claude Code, invoke /looper. To name the output folder, provide a name after the command; the README gives /looper client-onboarding-loop as an example.
Answer the interview and review the displayed ASCII flow preview before finalizing the generated artifacts.
Expected outcome
Looper creates a loop-design output folder with a preview and artifacts including loop.yaml, compiled loop.resolved.json, LOOP.md, RUN_IN_SESSION.md and a thin run-loop.py.
Guide limitations
- This documents installation and design, not successful execution of a generated loop.
- The installer creates a private .venv and installs PyYAML for the helper compiler.
- The README says the tool offers to run the loop in the same Claude Code session; this guide does not claim that run has been tested.
Sources & review
Source record details
- Source availability
- Active
- Sources checked
- 13 October 2026
- Original website title
- ksimback/looper · Official homepage title · retained in its source language
Related options
Claude Code /goal — An in-session execution primitive for persistent objectives; Looper designs and specifies a more structured loop that can be handed to it.
Claude Code /loop and /schedule — Scheduling mechanisms that re-fire a task or skill; Looper designs the workflow they can run rather than scheduling it.